HOME
QUICK SEARCH
AUTHORS
KEYWORDS
HELP
BUGSPHUSE US 2026 Papers - Index
281 papers. Source: PHUSE (Pharmaceutical Users Software Exchange) US Connect conference proceedings.
| # | Title× | Authors× | Abstract× | Keywords× | Extras | Pages |
|---|---|---|---|---|---|---|
| 1 | A Practical Approach to Causal Inference | Sherrine Eid, Laura Watson |
Any general introductory statistics course aims to leave students with one of the most important and foundational takeaways in the field: correlation does not imply causation. This leaves us with a rather glaring and consequential question: how can we infer causation? Such a question may be easiest to understand through a worked example. This paper outlines a practical framework for performing causal analysis on observational data, guiding users through confounder identification, model specification, and effect estimation. As a motivating example, we leverage this framework against real-world data from the U.S. National Health and Nutrition Examination Survey (NHANES) to estimate the causal effect of smoking cessation on long-term weight change | causal inference, data science, average treatment effect, treatment effect, potential outcomes, proc step, propensity score | source presentation source |
10 |
| 2 | A Proc Compare Enhancement Macro | Chris Weiner | PROC COMPARE is commonly used in clinical trial data validation for SDTM, ADaM, and TLF datasets, but its default outputs can be both difficult to interpret and incomplete, especially when investigating discrepancies. This enhanced macro offers significant improvements, including more readable reports, detailed summaries of non-matching ID records, and output datasets for unmatched and duplicate observations. It provides advanced comparative analyses, such as univariate statistics, percent missing, and distinct value comparisons for specified variables. String differences are clearly highlighted-even for nonprintable special characters-making it easier to spot subtle issues in long text fields. The macro generates comprehensive duplicate reports, indicating which variables differentiate duplicates and summarizing results in a one-record output dataset for easier downstream confirmation and tracking |
proc compare, proc step, out=, data cleaning, clinical trial, clinical data, sas studio, univariate analysis, variable list, cdisc | source presentation source |
12 |
| 3 | A Skeptics Guide to Using AI for Programming | Franklin E. Menius | Like it or not, "AI" is everywhere - on your phone, in your fridge, and even government systems. But is it a buzzword, a costly gimmick, or a genuine productivity tool? The truth is: it can be all of the above. Chances are your organization is already using AI or is planning to soon. So let this skeptic walk you through safely using AI for programming without compromising your skills, productivity, clinical data integrity, or sanity. This presentation will take a pragmatic look at the current AI capabilities with an emphasis on limitations, pitfalls, and practical applications in creating clinical outputs. Let's face it, we don't have to be enthusiastic about AI, but we do need to understand how to work with it | data integrity, clinical data | source presentation source |
13 |
| 4 | Accelerating Study Build Using AI Enabled CRF Design | Jeremy Zhang, Aimee Basile |
Industry benchmarks show that, on average, it takes 4 months from protocol approval to first patient, first visit for Phase 1 studies. For Phase 1 studies it is critical to reduce this front-end time to speed up the clinical development of drug products. The industry has been looking at ways to reduce this time with digitization and standardization of the protocol which helps decrease the time for protocol development and supports study design and build. To further decrease study build timelines we are piloting a GenAI solution to digest the protocol and Schedule of Activities (SoA) and create draft CRFs based on a standard library of data collection metadata, historical CRF used in previous studies and de novo generated CRFs. This first draft of CRFs from the SoA allows parallel development of CRFs and protocol and should decrease the study team review time for a savings of up to four weeks, reducing the timeline to three months | data validation rules, proc standard, excel | source presentation source |
7 |
| 5 | Accelerating Study Insights: Agile Data Visualisation Using R-Shiny | Lin Yuan, Heather Howell |
Interactive, real-time data visualization is transforming how study teams assess efficacy and safety. This project leverages R Shiny to convert raw datasets into actionable insights with agility and precision. The application converts raw efficacy data into structured analysis datasets and interactive visualizations, enabling study teams to explore efficacy trends rapidly and intuitively. Visualizations include Waterfall plots, Spider plots, Swimmer plots, Kaplan-Meier analyses, and Overall Response Rate tables, with listings also in spreadsheet format. For safety, a dedicated Adverse Event (AE) module derives TEAE TFLs from raw datasets, enabling interactive exploration of safety profiles. Integrating safety with efficacy provides a holistic view of outcomes and balanced evaluation. By using this tool, study teams reduce reliance on static reporting and streamline transformation. An agile approach allows iterative refinement without waiting for SDTM/ADaM processing | adverse events, adam dataset, dose, pd, swimmer plot, change from baseline, percent change from baseline, raw data, regulatory submission, clinical trial, efficacy summary, phase i, progression-free survival, waterfall chart, proc freq, data cleaning, data science, data governance, cdisc | source presentation source |
6 |
| 6 | Acquiring a Compound with Multiple Clinical Trials? Let's Get Ready for a Type C Meeting! Key Considerations for Preparing Its Regulatory Submission | Mathura P. Ramanathan, Kajal Tahiliani |
Pharmaceutical companies are constantly researching and developing drugs and medicines. In the US, the Food and Drug Administration (FDA) is the ultimate agency that approves the drugs, vaccines , devices and other therapies, to ensure that the new drugs are both safe and effective for their intended uses. Sponsors interact with the FDA requesting specific meetings to discuss their regulatory questions related to their drug development programs. It is very common that pharmaceutical companies acquire drug development programs that are in various stages with multiple clinical trials, for several strategic reasons. Therefore, sponsors will have a broad range of questions while they streamline these newly added compounds towards their regulatory submissions. Type C meetings provide an appropriate forum to seek FDA input on a variety of topics in drug development | clinical research, clinical trial, clinical data, regulatory submission, surrogate endpoint, data integrity, phase i, clinical programming, logistic distribution, data science | source presentation source |
12 |
| 7 | Across the Multiverse ... Statistically Speaking | Cralen Davis | In a clinical or other controlled trial, the goal is often to demonstrate the validity of a claim or effectiveness of a product to a governing or regulatory body using specific measures from collected data, generally referred to as endpoints. It is very common for a trial to have multiple endpoints or multiple tests involving 1 or more endpoints. However, formal testing of multiple hypotheses within a trial may result in multiplicity, the increased probability of falsely concluding that a result is statistically significant. Guidance from both the U.S. Food and Drug Administration (FDA) and European Medicines Agency (EMA) states that special consideration should be given to adjusting the alpha level, or probability of observing a false positive, to account for multiplicity. This paper discusses some of the methods and approaches that can be used to address multiplicity. Keywords: multiplicity, type I error, hypotheses, adjustment, p-value, critical value | significance level, type i error, null hypothesis, proc step, bootstrap | source presentation source |
4 |
| 8 | Adopting gtsummary at Scale: How Roche Built a Companion to gtsummary to Standardise and Simplify ARD-Based Reporting | Daniel Sjoberg | {gtsummary} is a widely used R package for creating publication-ready summary tables across diverse fields. While its flexibility supports broad adoption, organizations often require consistent formatting and domain-specific functionality to meet internal and field standards. At Roche, we developed crane, a companion R package to gtsummary, to streamline adoption and ensure compliance with company reporting requirements. {crane} provides a Roche-specific gtsummary theme that applies standardized defaults for table appearance, as well as custom functions for creating tables too specialized to reside in the general gtsummary package. Importantly, this framework supports the production of Analysis Results Data (ARD)-based outputs This approach lowers barriers to adoption within an organization and ensures high-quality, standardized outputs across teams | proc means, clinical trial | source presentation source |
3 |
| 9 | Advancing Precision Medicine and Regulatory Readiness: Data Harmonisation and Statistical Methods for Companion Diagnostics Bridging Studies | Ummul Vara Qurratul Ain | Companion Diagnostics (CDx) play a pivotal role in precision medicine by linking drug efficacy to diagnostic device performance and ensuring targeted therapies reach the right patients. Bridging studies are essential for demonstrating concordance between Clinical Trial Assays (CTAs) and candidate CDx assays providing evidence to support regulatory decision-making. This paper focuses on preparation of harmonized analysis datasets according to CDISC standards, role of supplemental negative and surrogate data in enhancing specificity and evaluating CDx performance relative to CTAs. Additionally, the framework describes statistical methods with a focus on clinical efficacy, concordance, and sensitivity analyses incorporating advanced approaches to address missing data in CDx studies and strengthen robustness of the results | sample, clinical trial, sensitivity analysis, imputation, bootstrap, confidence interval, adam dataset, multiple imputation, objective response rate, analysis set, data standardization, statistical programmer, missing value, primary endpoint, quality control, clinical programming, proc step, data cleaning, clinical endpoint, efficacy summary, proc surveyselect, sample size, data science, cdisc | source presentation source |
10 |
| 10 | Advancing Transparency, Patient-centricity, and Trustworthiness in Pharma: Metrics and Trends from the Good Pharma Scorecard | Jennifer Miller | Presentation | clinical trial, clinical data, clinical research | source | 48 |
| 11 | Advancing Transparency-by-Design in Clinical Trial Disclosure: Real-World Applications for Regulatory Submissions | Obaraboye Olude, Chantal Din-Bell |
Presentation | clinical trial, regulatory submission | source | 60 |
| 12 | AI + Metadata : An Agentic Approach to Standards-Based Automation | Kevin Burges | Clinical trial biometrics demand strict adherence to regulatory principles of Accuracy, Reproducibility, and Traceability (ART). However, the probabilistic nature of Large Language Models (LLMs) poses a challenge to integrating AI into GxP-compliant workflows. This paper presents a practical framework for introducing AI into clinical trial biometrics, addressing the real-world challenges of policy, regulation, trust, and technology. The framework combines a risk-based governance approach with technical strategies including metadata-driven AI, Model Context Protocol (MCP), and deterministic-first principles with human-verified probabilistic fallback. The discussion addresses the role of emerging CDISC standards - USDM, Biomedical Concepts, and Dataset Specialisations - in establishing the semantic framework necessary for reliable, autonomous agentic AI processes | clinical trial, confidence level, sas risk engine, data cleaning, cdisc | source presentation source |
10 |
| 13 | AI as Copilot: Empowering, Not Replacing, Clinical Statistical Programmers | Bhavesh Patkar | The rapid rise of artificial intelligence (AI) and large language models has sparked debate about whether programmers will be replaced by machines. In reality, AI is not a substitute for human expertise but a powerful copilot that can augment productivity, enhance quality, and reduce repetitive work. This paper explores how AI tools can support clinical statistical programmers in clinical research by assisting with knowledge retrieval, code generation for well-defined tasks, pre-validation quality control, and program understanding and reuse. Rather than displacing programmers, AI enables clinical statistical programmers to focus on higher-value activities such as complex analyses, problem solving, and ensuring compliance with regulatory standards. Practical examples demonstrate how AI can be integrated responsibly into clinical statistical programming workflows, highlighting both benefits and limitations | statistical programmer, clinical programming, data cleaning, clinical research, quality control, proc sql, proc step, clinical trial | source presentation source |
5 |
| 14 | AI Code Generator: Automating TFL Programming with ARS Metadata | Bhavin Busa, Navin Dedhia |
TFL programming has traditionally required extensive manual coding in SAS and R, consuming valuable time and resources. To address this, we developed the AI Code Generator, an LLM/GenAI -based solution that leverages CDISC Analysis Results Standard (ARS) metadata within the TFL Designer tool to automatically generate production-ready TFL code in both SAS and R. Early evaluations demonstrate more than 70% reduction in time for TFL output generation, underscoring the potential of AI -assisted programming to streamline clinical reporting workflows. The premise is simple: design and specify early in the process, with metadata and tools serving as the key drivers of efficiency. In this demo, we will present a live case study showing how table programs can be generated in minutes, quickly modified, and applied to study data to produce outputs, transforming the way statistical programmers' approach TFL development | clinical programming, data cleaning, clinical trial, statistical programmer, sas programming, data science, cdisc | source presentation source |
5 |
| 15 | AI is coming for you: New Biometric Leadership in the era of Agentic AI | Kevin Lee | Presentation | source | 36 | |
| 16 | AI Reading and Understanding the SAP to Generate the TFL TOC | Siqi Wang, Toshio Kimura |
What if AI can read and understand the SAP and learn about endpoints, analysis sets and analysis methods? Can AI then generate the TFL TOC? We developed a TOC generating AI solution using Python and ChatGPT. After uploading the SAP, AI agents collect all SAP components and generates the TOC. The solution consists of: 1) Document processing: Prepare AI-ready document 2) Metadata collection: AI agents collect SAP metadata 3) TOC generation: Using collected metadata, AI generates TOC We will demonstrate an AI solution generating the TOC. Key learnings include: 1) Large context: Processing large context documents such as the SAP requires a different more scalable AI strategy. 2) Specialized AI agents: We developed specialized AI agents specifically designed to collect different SAP metadata. 3) Hybrid AI and traditional programming: AI combined with traditional programming yielded the best results. If AI can generate the TOC … what else can it do? | data standard, data science, analysis set, secondary endpoint | source presentation source |
6 |
| 17 | AI Won't Replace Leaders-But It Will Redefine Them | Chris Decker | Presentation | data standard | source | 9 |
| 18 | AI-Assisted Metadata Programming: A Transparent Framework for PCS in Lab Data Analysis | Donghua Zeng | Statistical programmers often face the challenge of translating complex protocol or SAP text into thousands of lines of conditional code. This manual process is time-consuming, error-prone, and difficult to audit. Traditional approaches to deriving Potentially Clinically Significant (PCS) flags in lab data rely heavily on study-specific coding, which reduces consistency and slows delivery. This presentation demonstrates an AI-assisted, metadata-driven approach that streamlines programming while enhancing transparency and reproducibility. Free-text rules from protocols and SAPs are translated into Boolean criteria within structured metadata (EXCEL/TSV). Standard SAS or R macros then apply the metadata to automate PCS derivations. Programmers no longer re-code logic manually; instead, they review and refine metadata tables, ensuring traceability from text → metadata → code → results |
data cleaning, excel, call execute | source presentation source |
21 |
| 19 | AI-Augmented Bayesian Methods for Late-Onset Toxicity in Early-Phase Trials | Rachana Lele, Arfan Afzal |
predictive probability, neural network, dose escalation, dose | source presentation source |
8 | |
| 20 | AI-Driven Metadata Integration in Modern TLF Development | Chuck Gelb | audit trail, data cleaning, protocol amendment | source presentation source |
4 | |
| 21 | AI-Driven Quality Risk Management for Contract Manufacturing Oversight: An ML Approach | Kiran Kumar Gande | data cleaning, machine learning, gradient boosting, proc hpforest, feature importance, logistic regression, key performance indicator, quality control, neural network, stratified sampling, quality assurance, logistic distribution, regression analysis | source presentation source |
7 | |
| 22 | AI-Optimised AESI Flagging: Streamlined SAS Hash Table Logic with Dynamic CALL EXECUTE and Metadata-Driven Specifications | Hang Zhong, Rashad Rasul, Jiaqi Hu |
Traditionally, Adverse Events of Special Interest (AESI) workflows rely on standalone datasets and PROC SQL merges, leading to fragmented logic, performance bottlenecks, and high maintenance costs. This paper presents a metadata-driven approach that eliminates intermediate AESI datasets using SAS hash table logic and CALL EXECUTE for dynamic macro execution. A centralized AESI strategy specification replaces traditional lookup datasets, defining MedDRA strategies-Preferred Terms (PTs), High Level Group Terms (HLGTs), Standardized MedDRA Queries (SMQs), and customized SMQs (cSMQs)-for modular, reviewer-friendly traceability compliant with CDISC Analysis Data Model (ADaM) standards. Pilot implementation achieved an 85% reduction in AESI flagging effort (13 hours to 2 hours per study) with 100% concordance with traditional SQL methods, redirecting programmer capacity to higher-value safety analyses |
hash table, data cleaning, proc step, proc sql, lookup, excel, proc compare, call execute, where statement, audit trail, clinical programming, response variable, length statement, adverse events, macro definition, cdisc | source presentation source |
12 |
| 23 | AI-powered approach for faster, more efficient authoring of clinical and regulatory documents | Abhijeet Arora | In this paper, we discuss how an intelligent assistant powered by artificial intelligence (AI) can transform the authoring of clinical and regulatory documents such as Clinical Study Protocols, Clinical Study Reports (CSRs), Development Safety Update Reports (DSURs), Statistical Analysis Plans, and more. The pharmaceutical industry currently faces significant bottlenecks in document creation, often relying on manual transcription and static templates that are prone to error. By leveraging advanced AI models and Natural Language Processing (NLP), the proposed assistant adapts to user needs and acts as a digital partner under the guidance of subject matter experts. Unlike traditional manual authoring, this solution continuously learns from user interactions, improving both accuracy and efficiency over time | machine learning, if-then-else, clinical trial, data cleaning, data science, adverse events | source presentation source |
4 |
| 24 | AI-Powered Multiple-Agent Pipeline for Automating ADaM Dataset Generation | Bo Ci | adam dataset, quality control, data cleaning, error message, cdisc | source presentation source |
4 | |
| 25 | AI-Supported Validation of Open-Source Software | Sven Prasse | Programming in R for submissions is advancing rapidly. It is easy to find many high-quality specialized R packages contributed to and maintained on open-source platforms (e.g. Pharmaverse, Tidyverse). These packages are usable according to the statistical requirements but not ready for the regulatory requirements because of the lack of computer system validation. The R Validation Hub provides guidelines and tools, but R package validation is a repetitive, resource- and cost-intense company-specific task. AI is designed to solve repeating tasks based on textual descriptions. Consequently, entimo has embarked on a project using AI to support R package validation to thus save time and resources for clinical trial analysis | data cleaning, clinical trial | source presentation source |
5 |
| 26 | An AI-Driven Framework for Automated Statistical Analysis Plan Generation in Clinical Trials | Qingyuan Zhang | Statistical Analysis Plans (SAPs) are foundational documents in clinical trials, yet their creation is traditionally time- consuming, manual, and prone to inconsistency. We present an AI-driven framework that uses large language models (LLMs) to streamline and partially automate SAP generation from protocol documents. The system parses unstructured content to extract structured metadata, builds a clinical trial ontology, and applies reasoning logic to infer endpoints, analysis populations, estimands, and statistical methods. It adapts sponsor-specific SAP templates dynamically, generating high-quality, traceable sections that align with regulatory expectations. The framework balances automation with human oversight, enabling statisticians to refine and validate content through an interactive interface. This approach significantly reduces preparation time, improves consistency, enhances auditability, and ensures alignment between protocol intent and analysis plans | analysis set, data science, missing value, clinical trial, data cleaning | source presentation source |
8 |
| 27 | An RShiny App for Real-Time Continuous Safety Monitoring in Post-POC Trials: A Bayesian Approach to Decision-Making | Jinquan Wu | This paper presents an interactive RShiny application purpose-built for real-time, continuous safety monitoring in post-POC clinical trials. The app empowers clinical and statistical teams to conduct timely, data-driven safety assessments through an intuitive interface-eliminating the need for extensive coding or ad-hoc analyses. Key innovations include flexible data input, robust model configuration, and dynamic visualizations (text summaries, density plots, forest plots) that streamline the safety review process. Developed in accordance with FDA guidance [1], the application operationalizes Bayesian statistical methods [2,5] in a user-friendly environment, making advanced analytics accessible to trial teams and supporting proactive decision-making. Keywords: RShiny application, Continuous safety monitoring, Bayesian analysis, Real-time data visualization, post-POC trials, Decision support | clinical trial, relative risk, forest plot, posterior probability, adverse events, credible interval, density plot, bayesian statistics, likelihood function | source presentation source |
11 |
| 28 | Analysis Concepts Definition: Initial Perspectives from the CDISC Working Group | Stuart Malcolm | data science, clinical research, cdisc | source presentation source |
4 | |
| 29 | Anonymisation Strategies for Policy 0070 and PRCI Submissions | Stuart Donald | Presentation | proc means, clinical trial, threshold | source | 16 |
| 30 | Anonymization Insights from 2025 | Honz Slipka | Presentation | threshold | source | 11 |
| 31 | Applying Artificial Intelligence (AI) for Optimal Dose Selection in Oncology | Zoran Antonijevic | Traditionally, dose selection in Oncology trials was based on identifying the Maximum Tolerated Dose (MTD). This approach may lead to selection of a dose that results in excess toxicity without additional gains in efficacy. In 2021, the FDA launched Project Optimus, with objective of promoting more thorough assessment of dose response at early stages of Oncology development. This development paradigm, while not new in other therapeutic areas, is new to Oncology. It requires application of more sophisticated approaches, including innovative designs and clinical trial simulations. This paper explains how to utilize Artificial Intelligence (AI) to efficiently select the optimal dose through following steps: 1. Outline study designs and development strategies 2. Specify simulation scenarios 3. Execute simulation by using AI technology 4. Select the optimal strategy based on AI outputs | phase i, adaptive design | source presentation source |
5 |
| 32 | Around the Data DOSE-y Doe: Using DOSExx Variables within ADaM Datasets | Inka Leprince, Richann J. Watson |
In the realms of pediatric, oncology, and diabetic trials, a challenge in capturing treatment and dose levels arises when protocol-planned dose levels are adjusted during the treatment period. How can data analysts choreograph the Analysis Data Model (ADaM) data sets to capture these nuanced dose levels? The treatment variables TRTxxP/N in the Subject-Level Analysis Dataset (ADSL) and their partners TRTP/N in Basic Data Structure (BDS) and Occurrence Data Structure (OCCDS) are designed to group subjects by treatments as columns in the summary tables. But we also need to preserve the dose level adjustments on a subject- and record-level basis. DOSExxP and DOSExxA are used in ADSL, while their counterparts, DOSEP and DOSEA, lead in the BDS and OCCDS data sets. Together, these harmonious variables pirouette across the ADaM data sets, capturing the very essence of the adjusted dose levels in a dance that seamlessly unfolds | clinical trial, dose, treatment period, adam dataset, clinical research, clinical data, adverse events, data science, cdisc | source | 10 |
| 33 | AUTOCODE: Metadata-Driven Automation for Faster, Traceable SDTM and ADaM Programming | Matthew Finnemeyer | Statistical programmers increasingly maintain repositories of standardized SDTM and ADaM mapping specifications, sometimes within a metadata repository (MDR). This practice streamlines the creation of study-level specifications and promotes harmonization across studies, enabling future integrated analyses such as ISS and ISE. We enhance this approach through AUTOCODE, a process that parses human-language specifications into machine- readable syntax to automatically generate SDTM and ADaM dataset-creation programs; this reduces programmer time-on-task, strengthens traceability between specifications, metadata, and programs, and supports consistent implementation across studies. In this presentation, we demonstrate the AUTOCODE workflow, highlight key programming techniques for embedding automation into existing processes, and share lessons learned from implementation | raw data, adam dataset, cdisc | source presentation source |
6 |
| 34 | Automated Code Generation: Evaluating Deterministic and Probabilistic Approaches | Malan Bosman | In the last few years, the clinical development industry has experienced rapid changes with new CDISC standards (including the Analysis Results Standard), open-source packages, technology solutions, and Generative AI. As professionals in this industry, we need to make sense of the possibilities and risks, and forge new ways of doing things that are faster, cheaper, and of high quality. This paper unpacks different approaches for the automated generation of R or SAS programs producing Tables, Figures and Listings (TFLs). Two broad categories of approaches emerge: Deterministic and Probabilistic. The Deterministic approach refers to a system of logic and algorithms that produce consistent results for a given input, while the Probabilistic approach refers to a Generative AI model generating code with variability in the result. The pros and cons of both approaches are discussed based on demonstrations of each approach | adam dataset, analysis set, clinical research, data cleaning, cdisc | source presentation source |
6 |
| 35 | Automated File Comparison and Validation Across Programming Platforms | Darrell Edgley, Padma Kishan Korsapati |
Pharmaceutical companies periodically upgrade their integrated development environments (IDEs), requiring migration of code, macros, and data sets from older to newer systems. Additionally, legacy studies may be archived to free up storage for ongoing work. These transitions often involve bulk copying of files, which can be error-prone and time-consuming if validated manually. To address this, we developed a macro in SAS that automates the comparison and validation of files across programming environments. It ensures all files and macros are successfully transferred and checks for content consistency. For data sets, the macro uses PROC COMPARE to summarize differences, while discrepancies in other file types are highlighted in a text report. This approach significantly reduces manual effort, improves accuracy, and supports efficient migration and archival processes in regulated environments |
proc compare, proc step, sysrc, macro definition, flat file, macro variable | source presentation source |
16 |
| 36 | Automated Programming Consistency Checks: Enhancing Clinical Programming Quality, Efficiency and Scalability | Narendra Babu Kolisetty | Leveraging automation to improve accuracy and efficiency in clinical trial programming is vital for ensuring data integrity and regulatory compliance. This paper presents an innovative SAS macro-based solution that automates programming consistency checks by extracting and systematically analyzing macro parameter datasets across multiple programs. The utility macro applies predefined validation rules to identify programming issues such as missing ADaM and TLFs submission ready program files, incorrect subtitles, population and observation selections, therapy variables, AE percent incidence values in TLFs for DMC and CSR. Detected discrepancies are compiled into structured CSV files, with each tab corresponding to a specific check and containing detailed issue locations, affected output names, and ex planatory notes to facilitate rapid resolution. Automating this process significantly reduces manual review time and human error |
macro call, data cleaning, data validation rules, adam dataset, xls, macro parameter, regulatory submission, macro definition, excel, clinical programming, clinical trial | source presentation source |
5 |
| 37 | Automating Analysis Report Programming in SAS with VGSART: A Metadata-Driven Approach | Swathi Nattam, Piyush Makkapati |
The clinical programming industry is evolving, driven by the imperative to reduce costs and accelerate delivery timelines. To remain competitive, programming teams must go beyond and develop tools that drive efficiency ensuring consistency through innovation. One area consuming significant time is analysis output programming. VGSART is a metadata-driven SAS framework designed to streamline analysis outputs. It uses structured CSV metadata for table layouts, analysis variables, formats, titles and footnotes, and a single macro generates outputs ranging from simple summaries to c omplex shift tables. A detailed user guide lets contributors without prior TFL experience create accurate tables. In testing against QC -reviewed deliverables, VGSART cut programming time by 25% with fewer errors. The paper will cover framework design, adoption, macro flow, lessons learned, and outline plans to expand to additional output types |
mock shell, in=, shift table | source presentation source |
12 |
| 38 | Automating SDTM Compliance: A Web-Based SDTM Quality Control Tool Enhancing Clinical Trial Data Integrity | Indraneel Narisetty, Gopal Reddy Palvatla |
Accurate implementation of the Study Data Tabulation Model (SDTM) is essential for regulatory compliance and maintaining data integrity in clinical trials. We present a web-based SDTM Quality Control Tool that integrates standard industry checks with custom validation rules to identify domain-specific errors and cross- domain inconsistencies. Developed using Python and PYQT scripting, the tool automates validation processes, improves reproducibility, and ensures full adherence to CDISC standards. Additionally, it facilitates vendor oversight by generating clear, actionable reports that support timely issue resolution. This solution enhances the quality of SDTM deliverables and optimizes the quality control workflow, thereby accelerating reliable clinical data submissions | data cleaning, clinical trial, sdtm domain, data validation rules, clinical data, data pipeline, cdisc | source poster source |
11 |
| 39 | Automation vs AI: A Debate on Strategic Priorities in the Era of Data Science | Avinash Pati, Francis Kendall |
In an era of rapid technological advancement, statistical programmers face a pivotal question: should the primary focus today be on deepening automation capabilities or on advancing artificial intelligence (AI) adoption? This debate-style presentation brings together two expert perspectives to navigate this critical crossroads. One author contends that automation is the immediate priority, enabling programmers to streamline workflows, reduce human error, and free up valuable time for higher-level analytical tasks. They argue that robust automation lays the necessary groundwork for any future AI integration, ensuring consistency and scalability across programming environments. Conversely, the other author asserts that AI represents the future of statistical programming and that investment in AI- driven tools and methodologies must begin now | statistical programmer, clinical programming | source presentation source |
3 |
| 40 | AutoProgramming: Agents Are All Y ou Need | Chengxin Li | Statistical programming is resource-intensive and largely standardized with CDISC across the Biopharma industry, but programming codes not publicly shared with sponsor-owned intellectual property, limiting Large Language Model (LLM) training. While current LLMs struggle to produce fully reliable, production-ready code, they EXCEL in interpreting and processing documentation. To bridge this gap, this paper introduces a transitional multi-agent framework that integrates LLM-based AI Agents with SAS Macro Agents to enable scalable and reliable programming automation. In this framework, AI Agents interpret study documents to generate metadata that configures Macro Agent execution. For example, when provided with table shells and ADaM specifications, AI Agents extract Analysis Results Metadata (ARM) and display features, which Macro Agents then use to produce tables and executable SAS code |
clinical programming, macro definition, machine learning, clinical data, cdisc | source presentation source |
11 |
| 41 | Balancing Privacy and Utility with Smart Anonymisation | Zach Weingarden | Presentation | medical history, adverse events, high-level term, data mining | source | 13 |
| 42 | Batch Cleaning of SAS Programs for Submission Readiness: An AI and Python Hybrid Approach | Qiong Wei, Qisen Luan |
Preparing SAS programs to meet regulatory standards (e.g., clarity, conciseness and no irrelevant content, etc.) is critical in clinical trial submissions but often requires extensive manual effort. This paper presents a hybrid ML and Python-based approach for batch cleaning SAS programs to enhance code quality and submission readiness. The workflow combines regex-based comment extraction, alternative strategies for comment classification (from simple heuristics to ML prediction and hybrid ML + human review), Python-based batch cleanup with tracking logs, and post-cleanup validation by re-running programs and comparing outputs against the originals. This ML-assisted and human-in-the-loop approach offers a scalable solution for statistical programmers seeking to efficiently prepare high quality, submission ready SAS programs in clinical trial analysis | data cleaning, proc step, comment, proc compare, statistical programmer | source presentation source |
9 |
| 43 | Becoming a Lead Programmer: Skills, Responsibilities, and Career Pathways | Gregory Weller | Lead programmers are a critical part of biometrics teams that support clinical development. They are tasked with overseeing statistical programmers to ensure datasets and outputs are delivered on time, with accuracy, and to the highest quality standards. Moving from front-line statistical programmer (i.e. someone whose primary tasks are dataset and output programming) to lead programmer represents a key career milestone. Yet guidance on how to reach and succeed in this role is often limited. This paper will share practical insights into the transition to lead programmer, exploring what can be expected from the role both strategically and in day-to-day responsibilities. Key topics include: defining the core responsibilities of a lead programmer, identifying the skills and traits that drive success, and offering practical strategies for earning promotion into the role | statistical programmer, clinical programming | source presentation source |
5 |
| 44 | Beyond Administration: Leveraging AI for Strategic Clinical Project Management | Hayley MacKenzie | The dynamic landscape of clinical project management is undergoing a rapid transformation as artificial intelligence (AI) tools become increasingly integrated into daily operations. Platforms such as ChatGPT and Copilot are emerging as transformative partners in this space, offering capabilities that extend far beyond administrative support, enabling project managers to optimize efficiency, improve decision-making and strengthen overall collaboration. These technologies streamline core activities such as project planning, documentation oversight, and process automation, while also enhancing real-time communication, reporting, and cross-functional teamwork. By automating routine tasks, AI tools free project managers to focus on higher-value activities, including strategic oversight and stakeholder engagement | artificial intelligence, clinical trial, data cleaning | source presentation source |
7 |
| 45 | Beyond Box-Checking: Five Best Practices for Embedding QC in Modern SCEs | Chris McSpiritt | Quality control (QC) remains a cornerstone of regulatory confidence in clinical trial analyses, but too often it is handled as a separate, downstream activity. This paper shares best practices for integrating QC directly into statistical computing environments (SCEs), ensuring quality is built into workflows rather than added at the end. Drawing on lessons learned from global clinical teams and informed by industry survey data, we show how organizations can embed QC across data, code, and environments to reduce rework, strengthen reproducibility, and improve submission readiness. We present five concrete best practices: Git-first workflows, automation by default, flexible risk-based categorization, end-to-end collaboration models, and AI-ready architecture. By moving QC upstream and embedding it in daily practice, these approaches go beyond checklists to deliver continuous quality assurance in clinical analytics | statistical programmer, independent qc, audit trail, regulatory submission, data cleaning, adam dataset, clinical research, quality control, clinical programming, cdisc | source presentation source |
6 |
| 46 | Beyond the Hype: Practical Guidance for LLM Adoption in Statistical Programming | Michael Goedde | As organizations begin integrating Large Language Models (LLMs) into their data services, many struggle to move beyond experimentation and achieve reliable production use. This paper provides practical guidance for adopting LLMs in statistical programming environments, with a focus on embedded, task-specific systems rather than general-purpose tools. Drawing on real -world implementation experience, we discuss key challenges related to context management, performance, quality assurance, and data privacy. We further identify areas within the statistical programming lifecycle - such as controlled terminology mapping and SDTM variable mapping - where LLMs have demonstrated measurable and repeatable value. The goal of this paper is to support teams in adopting LLMs in a manner that balances innovation with operational rigor | raw data, clinical programming, controlled terminology, cdisc | source presentation source |
9 |
| 47 | Blind/Unblind Workflows: Secure Separation and Governance in the CDR SCE | Kala Shivali, Sachin Walawalkar |
Mishandled blinding jeopardizes studies. We present technical and process controls that keep treatment information separated, enforce entitlements, and prove segregation to auditors. The design includes split data products, dual pipelines, and redaction layers, with explicit approvals for unblinding events. We show how to prevent leakage and how to evidence that analysis code cannot infer assignment pre-lock. Attendees get a practical blueprint for building and operating blinding controls that satisfy statisticians and quality teams without paralyzing work. Keywords: Break Blind, Blind/Unblind | audit trail, interim analysis, data cleaning, inspection readiness, least privilege, access control, database lock | source presentation source |
11 |
| 48 | Breaking Free from the xpt: Exploration of Dataset-JSON as an Alternative Transport File to Regulatory Agencies | Benjamin Straub, Samir Parmar |
The SAS XPORT (XPT) transport file has been the de facto interchange format for SDTM and ADaM datasets in regulatory submissions for decades. While durable and widely supported, XPT is inherently constrained by legacy design decisions and is not a natural "first choice" format within modern open-source data ecosystems. The CDISC Dataset-JSON standard provides a contemporary alternative designed to represent clinical datasets and metadata using JavaScript Object Notation (JSON), enabling straightforward interaction across many programming languages and validation toolchains. R Consortium R Submission Pilot 5 explored a fully reproducible, publicly accessible submission-like package centered on R-generated ADaM datasets and delivery of datasets in Dataset-JSON v1.1 | adam dataset, regulatory submission, raw data, data cleaning, cdisc | source presentation source |
8 |
| 49 | Breaking into Statistical Programming: How SAS, R, Python, AI and ML Are Shaping Career Paths | Kishore Pothuri | Breaking into statistical programming without direct industry experience can feel daunting, but there is no single path to success in this evolving field. We spoke with statistical programmers from a variety of backgrounds, mathematics, technology, pharmaceuticals, and agronomy and discovered that transferable skills, technical acumen, and strong problem-solving abilities often outweigh formal credentials. A degree specifically in statistical programming is not essential; what matters most is analytical thinking, coding proficiency, and an understanding of clinical trial processes. Professionals are drawn to this field for its stability, career growth, and the opportunity to contribute to life-changing healthcare innovations | clinical programming, clinical trial, statistical programmer, clinical research, data science, machine learning, clinical data, regulatory submission, life sciences, data wrangling | source presentation source |
6 |
| 50 | Bridging Analysis and Strategy: The Transformative Role of Statistical Analysts in Drug Development | Lingsong Yun | clinical research, pharmaceutical industry, clinical trial, regulatory submission, csr, data science, imputation, sample, missing value, artificial intelligence, data integration, clinical programming, clinical data, sample size, sensitivity analysis, clinical endpoint, data cleaning | source presentation source |
17 | |
| 51 | Bridging Patient- and Trial-Level Insights | Catharina Dahlbo, Eva Kelty |
Behind every clinical data point is a patient, a person whose safety, comfort, and trust depends on how clearly we can see the full picture of their journey. Protecting patients requires moving beyond aggregated charts and static summaries to visualizations that bring together every relevant detail in context. In today's decentralized, data-rich trials, this means connecting clinical measures, operational activities, and patient-reported experiences into a coherent, patient level view. This poster will show how integrated, interactive visualizations can reveal the complete patient story, linking symptoms, lab trends, adverse events, and protocol deviations alongside lived experiences. Such a holistic perspective enables teams to recognize patterns early, act on subtle signals, and address risks before they escalate | adverse events, protocol deviations, data review, clinical trial | source poster source |
5 |
| 52 | Building Cross-Functional Regulatory and Programming Alignment for Successful Submissions | Ben Kaspar, Stephine Keeton |
Refuse-to-file (RTF) actions and technical rejection criteria (TRC) outcomes are often perceived by sponsors as last-minute crises arising during the 60-day filing review window. In reality, most filing failures are the predictable result of structural misalignment between strategy and programming teams centering on electronic submission requirements, and standardized study data implementation. While these issues typically arise years earlier in development they are frequently deferred past the point where it would be easy to address them. Typically, this is a structural problem of structure rather than scientific or regulatory issue. FDA law and associated guidance, including the RTF draft guidance, MAPP 6025.4, the Standardized Study Data Guidance, the Study Data Technical Conformance Guide, and eCTD guidance under section 745A(a) of the FD&C Act, clearly articulate expectations for completeness, organization, and technical conformity | phase i, data cleaning, controlled terminology, phase ii, cdisc | source presentation source |
6 |
| 53 | Building for the Long Haul: Managing Scope, Refactoring, and CI/CD in Internal R Packages | Blazej (Moni) Neradilek, Chenchen Yu |
Presentation | phase i, data wrangling, proc means, box plot, structured data, regression analysis, data cleaning, r package | source | 14 |
| 54 | Building High-Quality SDTM from Real-World Data: Lessons and Solutions | Peishan(Jenny) Han, Lixiao Chen |
Mapping real-world data (RWD) into CDISC SDTM presents significant challenges due to heterogeneous sources such as electronic health records (EHR), claims, and registries, as well as inconsistent or absent coding, nonstandard vocabularies, and variable data quality, etc. This work highlights key barriers and outlines pragmatic strategies to transform RWD into SDTM structures that meet regulatory and scientific requirements. We focus on critical considerations, including the application of inclusion/exclusion criteria, construction of study anchors (e.g., RFSTDTC) in the absence of protocol-defined visits, imputation of partial or missing dates, exposure reconstruction, AI-assisted line-of-therapy derivation, and resolution of critical issues identified through P21 validation. We also discuss best practices for building a submission-ready data package and explore how AI may facilitate lab unit harmonization, comorbidity adjudication, and treatment regimen standardization | baseline characteristics, imputation, clinical trial, regulatory submission, excel, data cleaning, sdtm domain, cdisc | source presentation source |
4 |
| 55 | Building Sustainable RBQM Frameworks: Challenges, Metrics and Real-World Results | Nicole Stansbury | Risk-based quality management (RBQM) is evolving from a theoretical approach into a practical framework to improve data integrity, patient safety, and resource utilization. This paper identifies remaining challenges, including embedding RBQM into established processes and securing adoption across functions. Proposed strategies to overcome these challenges involve workflow integration and cross-functional alignment. The paper also highlights key metrics that demonstrate the impact of RBQM in practice, such as the percentage of findings and protocol deviations (PDs) identified by central monitors, and the time taken to detect PDs by central monitors as compared to CRAs. A real-world case study of central monitoring assessment demonstrates how RBQM principles can be applied, the outcomes achieved, and the lessons learned. The paper aims to provide readers with practical guidance and actionable insights to strengthen oversight and advance the maturity of their companies' RBQM initiatives | protocol deviations, clinical trial, adverse events, data cleaning, raw data, concomitant medications | source poster source |
7 |
| 56 | Building the Information Infrastructure Required for AI with Human in the Loop | Tomás Sabat Stofsel | Presentation | hazard ratio, clinical trial, log-rank test, rank test, treatment effect, data cleaning, full analysis set | source | 53 |
| 57 | Business Event Triggers - The Next Level of Automation | Marc Jantke | Automation in clinical trial data management and analysis is traditionally designed as a set of pipelines: rigid sequences of steps, often initiated by manual input or by the completion of another pipeline. While effective for linear workflows, this approach lacks flexibility. Loopbacks, conditional branching, or constructing dynamic networks of activities are cumbersome or impossible. This presentation explores a new paradigm: automation through business event triggers. Every system activity produces events that can be captured and used to trigger subsequent actions. These events form a loosely coupled automation network rather than a fixed chain. Pipelines can still be mimicked where needed, but the event-driven model enables richer branching, context-sensitive reactions, and user-specific automations. Key benefits include handling diverse outcomes more elegantly, supporting complex workflows, and decoupling permissions | data transformation, clinical trial, raw data, clinical data | source presentation source |
6 |
| 58 | C&SP 203X Reinventing Clinical & Statistical programming for the AI - era | Sangeeta Bhattacharya | Presentation | clinical programming, data review, data cleaning, model validation, clinical data, database lock, clinical research | source | 19 |
| 59 | Case Study: Leveraging CDISC CORE for Proactive SDTM Compliance and Submission Readiness | Shivani Gupta | Ensuring compliance with CDISC standards is essential in preparing clinical trial datasets for regulatory submission. Traditionally, validation using commercial tools is performed only after all Study Data Tabulation Model (SDTM) domains are developed, oft en leading to late discovery of issues and costly rework. This paper describes our experience integrating the CDISC Open Rules Engine (CORE), an open- source validation tool, directly into the SAS environment to run compliance checks during SDTM dataset development. By embedding CORE into the workflow, we identified and resolved conformance issues earlier, improved data quality, and streamlined the overall submission process. We also compare outcomes of per-domain validation versus study-level validation, demonstrating how early checks reduce errors, accelerate timelines, and encourage a compliance- first culture | data cleaning, controlled terminology, clinical programming, sas base, macro variable, sdtm domain, cdisc | source presentation source |
6 |
| 60 | CCI Proposals In Clinical Documents - Gap Analyses | Radu Popescu, Silvia Garrido-Lestache |
Presentation | liquidity gap, sample size, sample, clinical research | source | 14 |
| 61 | CDISC 360i Demo: Automating from Study Design to Submission | Darren Moreland | phase i, clinical data, sdtm domain, data cleaning, phase ii, cdisc | source presentation source |
6 | |
| 62 | CDISC 360i: From Vision to Implementation - Lessons Learned and the Road Ahead | Peter Van Reusel | Presentation | sdtm domain, raw data, annotated crf, clinical trial, phase i, protocol amendment, excel, cdisc | source | 23 |
| 63 | CDISC ARS Template Code: Gearing Up with Automation in TFL Programming | Malan Bosman, Anna Yaggi |
With the advent of the CDISC Analysis Results Standard (ARS), TFL automation is being re-imagined. The concept of using structured metadata and generating results in an Analysis Results Dataset (ARD) format first, allows for new ways to automate TFLs. One particularly exciting opportunity for automation using ARS, is utilizing the "Analysis Method Code" concept. In this PHUSE 2025 paper, Richard Marshall referred to it as "Gear 5" of TFL Automation. It allows for macro-like code to be set up for each Analysis Method (e.g. frequency counts, or summary statistics, etc.), along with parameters sourced from the ARS metadata. In an R implementation, the 'CARDS' and 'cardx' packages (both part of 'pharmaverse' offer uniquely efficient functions for this use case. This presentation will dive into how these functions are used as template codes within the ARS metadata, and how we can create dynamic code to generate reliable, automated results for TFLs |
proc means, clinical research, cdisc | source presentation source |
7 |
| 64 | CDISC Biomedical Concepts for Breast Cancer: A CDISC 360i Perspective | Bess LeRoy, Brian Harris |
Biomedical concepts serve as the essential semantic framework that connects and aligns various standards across clinical research. These concepts provide a common language for representing critical study data, ensuring consistency, accuracy, and interoperability across systems. By linking key standards such as USDM, SDTM, and ADaM, biomedical concepts enable seamless data flow and support regulatory compliance across diverse platforms. This paper explores the vital role of biomedical concepts in advancing data harmonization in the context of the CDISC 360i, an initiative designed to close gaps in data standards by adding the required semantic layer to enable end-to- end automation. Using a breast cancer use case readers will gain insights into how biomedical concepts are being applied to enhance data exchange and foster cross-disciplinary collaboration, paving the way for more efficient and transparent clinical research | clinical research, clinical data, cdisc | source presentation source |
2 |
| 65 | Challenges and Solutions for Generating Synthetic CDISC Data | Sundaresh Sankaran, Catherine Briggs |
machine learning, correlation, life sciences, data cleaning, data science, data integrity, cdisc | source presentation source |
6 | |
| 66 | Changing rules on Data Ownership and Data Access | Darshan Kulkarni | Presentation | artificial intelligence, service provider, csr, adverse events, clinical trial, data protection, data pipeline | source | 38 |
| 67 | Check Yourself: Edit Checks and Queries Made Easy with Shiny | Maya Gans | Effective data management is critical in clinical trials to ensure clean, accurate, and compliant datasets. Shiny allows us to build applications tailored for clinical data management and enables CDMs to move beyond static listings and spreadsheets. By integrating edit check logic, discrepancies and flags can be raised in real time with minimal manual effort. Automated queries generated based on predefined rules/dynamic thresholds supporting consistent and efficient data cleaning. Customizable filters, dril l-down views, and responsive visualizations make it easier for CDMs/DMs to explore key issues across subject-level and study-wide datasets. To ensure performance and usability, strategies such as modular design, reactive programming, and server -side processing are employed to ensure smooth performance with large clinical datasets. Features such as audit logging and built-in metrics enhance traceability and oversight | clinical data, data review, data governance, data validation rules, clinical data management, clinical trial, access control, data cleaning, outlier, audit trail, raw data | source presentation source |
8 |
| 68 | Clinical Programming in the Fast Lane: Accelerating with SAS | Kim Peplinski | The demand for faster, more reliable clinical trial data analysis continues to grow as organizations face pressure to accelerate regulatory submissions without compromising compliance. SAS Clinical Acceleration provides an integrated statistical compute environment designed specifically for clinical research, combining automation, traceability, and scalability. This session will highlight how SAS Clinical Acceleration enables teams to streamline workflows, automate statistical programming jobs, and ensure reproducibility across studies. With built-in auditability and seamless integration of both SAS and open-source tools, the platform empowers clinical programmers to manage complex analyses while maintaining full regulatory compliance. By unifying data, code, and results in a controlled environment, SAS Clinical Acceleration reduces operational risk, shortens development cycles, and supports sustainable, future-ready clinical programming practices | sas studio, clinical programming, audit trail, sas viya | source presentation source |
7 |
| 69 | CLIQ: An AI-Native Framework for Unified Clinical Trial Data Quality Assurance | Gopinath Viswanathan | Clinical trial data validation is essential for ensuring accuracy and compliance in regulatory submissions. Current practices are fragmented across SDTM standardization, ADaM derivation, TFL generation, and CSR compilation, relying on siloed quality-control processes that are resource-intensive and error-prone. Traditional rule-based validation methods cannot capture cross -domain dependencies, allowing subtle inconsistencies to propagate undetected. This paper presents CLIQ (Clinical Intelligence for Qualit y), an AI-Native framework that unifies clinical trial data validation using graph neural networks (GNNs) and semantic reasoning. CLIQ models clinical trial data as knowledge graphs with a dual -decoder architecture for anomaly detection and repair suggesti ons | data cleaning, clinical trial, clinical data, quality assurance, adverse events, cdisc | source presentation source |
5 |
| 70 | Closing the Gap: Identifying Missing Tumour Assessments in Oncology Studies | Ke Xiao | The definition and identification of "missing two or more consecutive tumor assessments" (Missing2TAs) is a key consideration for censoring in progression-free survival (PFS) analysis in oncology clinical trials. However, operationalizing this concept can be challenging, especially in the presence of unscheduled assessments and protocol deviations. This paper introduces and compares three methods for identifying Missing2TAs: a visit-based method that utilizes mapping to protocol defined scheduled visits, a date-based method that relies solely on assessment intervals, and a hybrid method that combines visit mapping with assessment dates. Each approach is described in detail, with practical examples illustrating their applications. The paper aims to provide a clear framework for statistical programmers and clinical teams to select and implement the most appropriate method that fits their study context | protocol deviations, randomization, clinical trial | source presentation source |
8 |
| 71 | Code Switching: Parallels Between Human Languages and Multilingual Programming | Danielle Stephenson, Laura Mino |
As multilingualism gains importance in the pharma industry, we've observed that the processes for learning programming languages closely mirror those for learning human languages. What insights can we draw from the study of natural languages to enhance our approach to multilingual programming? What similarities can we find between a SAS programmer learning R and a native Mandarin speaker learning English? Join us, two polyglots who enjoy both people-talk and code-talk, as we explore natural and programming languages to see how techniques for understanding one can be leveraged to understand the other | if-then-else, pharmaceutical industry | source presentation source |
9 |
| 72 | Coding Clinical R with Confidence: Digital AI Assistant based on Self-Hosted LLMs | Kostiantyn Drach | Programming in R is gaining momentum in the clinical domain (e.g., Pharmaverse initiative). The rise of AI-powered tools is accelerating this trend, yet there are currently no well-known pre-trained LLMs specifically designed for this domain. To close this gap, we develop a self-hosted digital AI assistant built on coding LLMs (CodeLlama, StarCoder2, etc.), with expert-driven customization and evaluation. We leverage our over-decade experience in training clinical programmers globally to transform it into AI training workflow, targeting industry standards and predictable quality of code produced. Self-hosted models ensure privacy and full control across all stages, making them suitable for regulated environments. The framework combines retrieval- augmented generation (RAG), supervised fine-tuning (SFT), and continuous expert oversight to achieve quality of outcomes, controlled, verified, and benchmarked through multiple layers (human experts, specialized LLMs, etc.) | sample, clinical programming, adverse events, training set, proc freq | source presentation source |
8 |
| 73 | Combining Trigonometry, Functions and the Polygon Plot to Create Complex Figures Not Available in SAS, Including Sankey, Sunburst and CIRCOS Charts | Jeffrey P. Meyers | SAS currently has an expansive array of plot types available within the SG procedures and Graph Template Language, but there are still charts that are too complex to create using conventional means. Fortunately, SAS has a powerful trick up its sleeve to allow the ambitious programmer to design these charts: the polygon plot. A polygon plot allows the programmer to use geometric, trigonometric, and other equations to manually produce x and y coordinates that form the perimeter around a shape with the option to fill in the shape with color. This paper will walk through utilizing the polygon plot in combination with mathematical functions to create circular, twisting, and other complex graphs including the Sankey, sunburst and CIRCOS charts | bar chart, do loop, macro variable, series plot, proc freq | source presentation source |
16 |
| 74 | Common Issues in BIMO Clinical Site Dataset Packages | Michael Beers | When preparing a submission package for the Bioresearch Monitoring (BIMO) Clinical Site (CLINSITE) Dataset, it is important to ensure compliance to the FDA's specifications and guidance. It is often the case, however, that issues exist with the CLINSITE dataset and the associated documentation. This paper will review some of the most common issues seen across the industry, and how the issues should be addressed | safety population, treatment arm, adam dataset, missing value, data cleaning, adverse events, clinical trial, cdisc | source presentation source |
7 |
| 75 | Comparing Medical Queries with Interactive Graphics using JMP Clinical | Chris Kirchberg | adverse events, system organ class, excel, concomitant medications | source presentation source |
7 | |
| 76 | Complex Trials, Simple Disclosures: Streamlined Transparency of Innovative Study Designs | Ritama Gupta Dempsey | Presentation | clinical research, clinical trial, primary endpoint, protocol amendment | source | 15 |
| 77 | Consistency, Speed and Satisfaction: The Impact of TLF Mock Shell Centralisation | Jie Yang | In clinical trials, generation of tables, listings and figures (TLF) is traditionally guided by study-specific TLF mock shells derived from the statistical analysis plan. While well established, this approach can be resource-intensive and repetitive. To address these challenges, we implemented a central TLF mock shell library improving efficiency of TLF development. A central TLF mock shell library provides reusable standardised templates promoting consistency in TLF formatting and structure. It covers commonly repeated analysis domains (e.g. demography, adverse events) while remaining flexible to accommodate study-specific endpoints and nuances. This approach streamlines study set-up, reduces programming and validation effort, accelerates timelines and enhances consistency across studies. In this presentation, we will share our journey of transitioning from study-specific TLF mock shells to a central TLF mock shell library | mock shell, data cleaning, adverse events, clinical trial, data science | source presentation source |
6 |
| 78 | Creating Synthetic Data that Mimics Real-World Data for Accurate Analysis Programming Development | Linshan Yuan, Weiwei Xu |
Development of analysis programs in clinical trials often requires realistic test data before final study data becomes available. This paper presents effective methods for generating synthetic Anti-drug Antibody (ADA) data that preserves complex internal logical relationships. We developed a comprehensive approach utilizing the Synthetic Data Vault (SDV) framework combined with custom constraint implementations to generate realistic ADA datasets that maintain statistical properties and logical dependencies inherent in real-world clinical trial data. Our methodology demonstrates how synthetic data can significantly accelerate Tables, Figures, and Listings (TFL) generation while ensuring minimal code modifications when transitioning to actual study data. The approach is validated across multiple datasets, showing consistent preservation of complex relationships including conditional dependencies, monotonic constraints, and derived variable relationships | clinical trial, data cleaning, clinical data, data validation rules | source poster source |
5 |
| 79 | Data Can't Do It Alone: The Untold Power of Soft Skills in IDMCs | Maira Pinto | clinical research, missing value, adverse events | source presentation source |
9 | |
| 80 | Data Cleaning and Gene Expression Visualisation in Haematological Conditions | Nasreen Bano, Ashwika Papasani |
A new technology that offers a view into gene expression at the individual cell level or across the entire cell population is RNA sequencing. In clinical research, it is becoming increasingly crucial for diagnosing diseases and understanding how patients respond to treatment. In preclinical studies, this approach has already shown promise in assisting researchers in understanding the expression of genes and the functioning of cells. Large volumes of data are produced by this technology, necessitating the use of specialized bioinformatics tools to identify, clean, and eliminate low-quality cells, group cells according to gene expression, and produce visualizations that show variations. In order to investigate blood cell formation, we examine in this presentation how this data is collected, processed, and visualized | correlation, quality control, sample | source poster source |
6 |
| 81 | Data Designer: Empowering Data Preparation, Transformation, ARD Creation and Summarization for Insights and Decision Making | Mukul K Mittal, Sanjeev Rathore |
Data Designer is intuitive R Shiny application, built to streamline and accelerate data transformation, as well as to summarize trials data. The tool supports seamless ingestion of datasets from various file formats e.g., SAS, XPT, CSV, EXCEL, Parquet, and JSON. At its core, Data Designer empowers users to create Analysis Results Datasets (ARDs) using {card} package by leveraging analysis results Metadata specs. These ARDs can be further summarized using {gtsummary} to generate publication ready clinical tables . Application also offers support for multilingual outputs and includes features for generating mock tables from dummy data as well. For customized reporting, users can create summary outputs from the ADaM datasets using {rtables}, either through predefined templates or custom logic, with outputs exportable in RTF and DOCX formats |
missing value, imputation, adverse events, tables, excel, clinical trial, cdisc | source presentation source |
8 |
| 82 | Data Diaries Don't Lie: Stats That Spot ePRO Risks | Steven A. Gilbert, Nechama Katan |
Electronic Patient-Reported Outcomes (ePRO) are increasingly used in clinical trials to support key efficacy and safety endpoints. Because ePRO data are entered directly by patients, traditional approaches such as record-level review and query management are often not appropriate to confirm that the data are fit for purpose. Sponsors remain responsible for ensuring data integrity, and audit trail metadata provide a practical way to evaluate how ePRO data are generated and whether observed behavior is consistent with independent patient entry. This paper describes statistical methods for using ePRO audit trails to identify data integrity risk signals, with emphasis on timing- and behavior-based metadata such as open and close times, device identifiers, and masked PIN codes. Common risk scenarios include delayed or clustered entries, unusually fast completion times, data entered at unusual hours, and patterns that suggest site staff may be entering data for patients | audit trail, normalization, proc means, root cause analysis, trend analysis, clinical trial, outlier | source presentation source |
12 |
| 83 | Data Harmonisation for ISS and ISE: A Practical Checklist for Integration Success | Qiong Wei, Lily Cheng |
Integrated Summary of Safety (ISS) and Integrated Summary of Efficacy (ISE) are pivotal components of regulatory submissions, requiring harmonized, CDISC-compliant data across multiple clinical studies. Having a practical checklist helps guide programmers through the integration process, addressing common challenges such as alignment of dataset implementation guides (IGs), harmonization of variable attributes, formats and casing, reconciliation of measurements (e.g. --PARAM/--PARAMCD) and units, and standardization of coding dictionaries (e.g. MedDRA, WHODRUG, LOINC) and controlled terminology, alignment of visit mapping and derivation logic of key analysis variables (e.g., ABLFL, ANLxxFL, AVALCATx), and coordination of subject identifiers for subjects participated in multiple studies. By applying this checklist, teams can improve integration accuracy and ultimately strengthen the overall integrity of ISS and ISE submissions | controlled terminology, label statement, data cleaning, vital signs, adverse events, adam dataset, dose, data science, study day, character variable, clinical trial, regulatory submission, cdisc | source presentation source |
10 |
| 84 | Data Storytelling: A Crucial Leadership Skill in Today's Clinical Research Landscape | Bhavesh Patkar | In an era of increasing data complexity, the ability to tell a clear and compelling story with data has become a core leadership skill. While technical expertise ensures accurate analyses, it is effective data storytelling that bridges the gap between data and decision-making. Leaders who can translate numbers into meaningful narratives drive alignment, influence stakeholders, and accelerate outcomes. This paper explores data storytelling as a critical capability for today's biometrics leaders. It highlights how to frame data within context, tailor messages for diverse audiences, and balance technical accuracy with accessibility. Through illustrative examples and practical strategies, the paper demonstrates how effective data storytelling can transform clinical data into actionable insight, enabling leaders to communicate with clarity, build trust, and drive meaningful impact in a data-driven clinical research environment | data storytelling, clinical research, adverse events | source presentation source |
9 |
| 85 | Deceptive Twins: An Interactive Python Tool for Automated RTF Document Comparison in Clinical Submissions | Aravind Gajula | In clinical trial programming, routine updates ranging from a minor footnote edit to a full dataset refresh can trigger wholesale regeneration of tables, listings, and figures. While the underlying content often remains unchanged, refreshed timestamps and formatting requirements force reviewers to re -examine outputs line by line to confirm that nothing meaningful has shifted. As documents progress through multiple regulatory revisions, this problem intensifies, giving rise to what we call "deceptive twins": files that look virtually identical at first glance, yet may conceal subtle bu t critical differences. To address this challenge, we developed Deceptive Twins, an interactive Python application that automates the comparison of Rich Text Format (RTF) files commonly used in clinical deliverables | clinical trial, regulatory submission, clinical programming | source presentation source |
8 |
| 86 | Defining Time Zero: Evaluating Index Line Selection Methods in External Control Arms (ECAs) | Peishan(Jenny) Han, Lorraine Fang |
Establishing an external control arm (ECA) for single-arm trials requires defining the index date (time zero), the point at which a patient would be eligible in a contemporary study. In real-world data (RWD), patients often have multiple qualifying lines of therapy (LOT), creating complexity in index-line selection. We evaluated several approaches, including First Eligible Line, Last Eligible Line, Second-to-Last Eligible Line, All Eligible Lines, Restricted to 3rd Line, Stratified Random Line and Random Eligible Line. Using real study data, we examined the impact of these methods on treatment-effect estimates for a target therapy versus standard of care (SOC). To our knowledge, this is the first study to compare index- date assignment strategies in RWD using actual study data rather than simulations, providing important insights for the design and interpretation of ECAs | sample size, sample, treatment effect, clinical trial | source presentation source |
5 |
| 87 | Democratising SDTM Transformations: AI-Assisted Function Development with Business-Driven Governance | Ulf Nielsen, Rostislav Markov |
statistical programmer, study day, data cleaning, clinical programming, quality assurance, clinical research, cdisc | source presentation source |
11 | |
| 88 | Demonstrations of AutoTable | Chengxin Li | This paper presents AutoTable, an automated analysis -generation toolset implemented entirely in SAS macros, where parameters are aligned with Analysis Results Metadata (ARM) fields. The AutoTable system presented here focuses on efficiently and reproducibly generating non -inferential analysis tables, integrating automation directly into existing statistical-programming workflows. Four core macros have been developed -%ADSLEVL, %BDSSTAT, %BDSSHIFT, and %OCCFREQ-each supporting a different category of non-inferential analysis. For every macro invocation, the following outputs are automatically generated: Table (RTF , PDF , or XLSX formats) Plain, executable, submission-ready SAS code Final result dataset (RDS) Associated metadata (ARMDS), facilitating define-xml generation * Log file During the live demonstration, four representative call instances will be showcased: 1. Demographic table using %ADSLEVL. 2. SOC/PT/Severity table using %OCCFREQ 3 | macro call, macro definition, shift table, adam dataset, clinical programming, cdisc | source presentation source |
11 |
| 89 | Demystifying TMLE: A Practical Approach for Clinical Data Scientists | Joshua Enxing | Targeted Maximum Likelihood Estimation (TMLE) is transforming causal inference in clinical data science by pro- viding reliable effect estimates from complex observational and trial data. Traditional methods can produce biased or imprecise results, particularly when study designs are imperfect or confounding exists. TMLE addresses these challenges through a doubly-robust framework: if either the outcome model or treatment model is correctly spec- ified, the estimator remains consistent, and if both are correct, it achieves semiparametric efficiency. Despite theoretical complexity, TMLE can be implemented practically using accessible tools. In this talk, we break down the TMLE framework for a clinical data audience, showcase studies where it improved precision and reliability over classical methods, and provide guidance on best practices for applying TMLE to real-world datasets | regression analysis, logistic distribution, odds ratio, clinical trial, maximum likelihood estimation, likelihood function, randomization | source presentation source |
12 |
| 90 | Deriving Phantom Records in BDS Datasets To Support the FDA's Standard Safety Figures for Missing Data Analysis | Guowei Wu, Ruchitbhai Patel |
Missing and existing data analyses are key components of safety evaluations for laboratory parameters in NDAs and BLAs, as outlined in recent FDA guidance on standard safety tables and figures. Handling missing data is a critical challenge in clinical trials, potentially impacting study validity. The Basic Data Structure (BDS) dataset, a core part of CDISC ADaM standards, provides a robust framework for creating analysis -ready datasets. This paper presents a systematic approach to derive phantom records within BDS datasets to represent missing observations implied by study design or visit schedules. Incorporating these phantom records enables analysts to better characterize missing data patterns and conduct sensitivity analyses. The method leverages protocol -defined visit windows and data relationships to infer and generate phantom records | missing value, analysis set, sensitivity analysis, randomization, cdisc | source presentation source |
6 |
| 91 | Design of Everyday Shiny Apps | Maya Gans, Casey Aguilar-Gervase |
data science, error message | source | 5 | |
| 92 | Detecting Trends Before They Appear: AI-Driven Visualisations in Clinical Trials | Steven Tan | clinical trial, data visualization, resource allocation, clinical data | source presentation source |
6 | |
| 93 | Deterministic Foundations, AI Acceleration: Lessons from Developing Hybrid Clinical Review Tools | Matthew Kumar, Srinivas Veeragoni |
As large language models (LLMs) gain traction in clinical development, teams face a fundamental question: when should deterministic programming suffice, and when can probabilistic AI accelerate insight generation? We developed a medical review R Shiny application using a hybrid approach that combines robust, deterministic components for core analyses with AI-assisted exploration that accelerates medical reviewers' ability to address ad- hoc, novel questions with flexibility and velocity. This development journey revealed critical trade-offs-speed versus reproducibility, flexibility versus consistency, and breadth versus control-while highlighting the importance of transparency, human oversight and building (and maintaining) end user trust. We discuss practical consequences of the hybrid choice and propose suitability criteria based on problem definition, risk, criticality, and required traceability | clinical research, data review, scatter plot, waterfall chart, data cleaning, proc sql, clinical trial | source presentation source |
8 |
| 94 | Developing the Workforce of the Future | Avinash Pati | In an era defined by rapid technological innovation and data-driven-decision-making, statistical programmers (SP) occupy a pivotal role in shaping analytical pipelines. This paper provides strategic guidance to SP seeking to advance their professional trajectories and maximize their impact in the evolving analytics ecosystem. Outlining practical steps for upskilling, emphasizing the importance of mastering foundational languages such as R/Python/SAS and emerging technologies including machine-learning-frameworks/cloud-based-tools/AI-applications. The paper delves into the value of building interdisciplinary communication skills, engaging in open-source-communities, and pursuing relevant certifications. Special attention is given to navigating the shift from legacy ways to modern, collaborative platforms, highlighting opportunities for leadership in adopting reproducible research practices and automation | statistical programmer, regulatory submission | source presentation source |
5 |
| 95 | Development Process of the Bioresearch Monitoring (BIMO) Package and the Obstacles Successfully Addressed | Sarad Nepal, Murali Neela |
Presentation | adam dataset, clinical trial, clinical data, data audit, data integrity, missing value, proc report, cdisc | source | 19 |
| 96 | DevOps for Clinical Data Science: A Leap into the Future of Clinical Analysis | Frikkie Oppel, Riaan Van Straaten |
Bioforum is modernizing its clinical programming environment by transitioning from traditional, folder-based workflows toward a dynamic Clinical Data Science Model built on Azure DevOps and Git. This transformation aims to strengthen governance, automation, traceability, and audit readiness across clinical analysis activities. This paper outlines the objectives and operating model used in this transition, emphasizing computerized version control, structured branching strategies, automated CI/CD-based validation, comprehensive audit logging, and controlled management of blinded versus unblinded activities. The approach is aligned with regulatory expectations, including 21 CFR Part 11 for electronic records and electronic signatures, ICH E6 (R2/R3) for Good Clinical Practice, and ALCOA+ data-integrity principles | 21 cfr part 11, clinical programming, data cleaning, access control, audit trail, data science | source presentation source |
6 |
| 97 | Digital Twins for Clinical Submissions: Simulating Regulatory Review Before Submission | Heather Gorr, Divya Kapoor |
Regulatory submissions in the life sciences are often delayed by unforeseen reviewer queries, inconsistencies in metadata, or gaps in documentation quality. This paper introduces a "digital twin" framework designed to replicate the end -to-end regulatory review process before submission. Leveraging AI-driven validation, reproducibility checks, and simulated mock inspections, the system enables sponsors to proactively stress -test datasets, outputs, and narratives against regulatory expectations. Unlike conventional quality control methods, the digital twin model provides a dynamic feedback loop that highlights inconsistencies, flags non -compliance risks, and offers targeted remediation strategies before database lock. Early case studies demonstrate a measurabl e reduction in submission-cycle iterations and query volumes, alongside improved reviewer confidence in data integrity | regulatory submission, data cleaning, clinical trial, quality control, machine learning, artificial intelligence, database lock | source presentation source |
9 |
| 98 | Electronic Clinical Data Submissions Across Global Regulatory Authorities | Sampath Madanu | Electronic submissions to global regulatory agencies, such as the FDA, EMA, PMDA, Health Canada and NMPA, are necessary to get approval for medicinal products. These agencies have different requirements for submissions. The FDA, EMA, and PMDA require technical compliance with the ICH's eCTD, but some regulators may have additional requirements. It is essential to contact regulatory agencies prior to submission to learn about their specific requirements. The FDA and PMDA mandate data submission in a standardized CDISC format, including SDTM for collected data and ADaM for analysis datasets. The PMDA strongly recommends including ARM, while it is optional for the FDA. Knowing these key regulatory requirements is important for planning, creation, and submission of clinical study data | clinical trial, clinical data, data cleaning, regulatory submission, csv, adam dataset, sdtm domain, statistical programmer, cdisc | source presentation source |
12 |
| 99 | Electronic Submission of Population PK Analysis Files ... What's Your Modus Operandi? | Luai R. Alzoubi | Population Pharmacokinetic (PopPK) analyses play a critical role in drug development and regulatory decision making. FDA has emphasized the importance of reproducibility in PopPK submissions by highlighting the value of these analyses and the inclusion of data and related files (e.g. coding scripts) in submission packages. All conclusions drawn from the PopPK analysis should be reproducible by the Agency with submitted codes and data. However, submission practices vary across sponsors. Key challenges include integrating heterogeneous data sources, managing sparse PK sampling, and ensuring consistency across studies. Additionally, submission challenges such as aligning documentation formats, meeting evolving regulatory expectations, and ensuring traceability. Pharmacometrics Analysis Report (PMAR) dataset programming plays a vital role in overcoming these issues | regulatory submission, population pk, clinical trial, dataset, adam dataset, pd, key, data wrangling, data cleaning, clinical research, concomitant medications, cdisc | source presentation source |
13 |
| 100 | Empowering Clinical Data Review with Interactive Patient-Level Visualisations in {teal} | Dony Unardi | data review, clinical data, clinical trial, data cleaning, clinical research | source | 4 | |
| 101 | Enabling Innovation: Lessons from the CDISC AI Innovation Challenge | Julie Smiley | As artificial intelligence becomes an integral part of clinical research, standardized, interoperable data is essential to unlock its full potential. In this panel session, we showcase insights and outcomes from CDISC's first AI Innovation Challenge, a global call to action inviting technology innovators to build AI/ML solutions that use or accelerate development of CDISC standards. Panelists will reflect on what worked and what didn't, from data accessibility to the maturity of AI approaches, and discuss lessons learned for future innovation. The session will also examine insights gleaned from the submitted use cases: where value was generated, what patterns emerged across solutions, and how these technologies might reshape study design, data transformation, and regulatory workflows | clinical research, clinical data, data cleaning, cdisc | source presentation source |
10 |
| 102 | End-to-End Workflow for CDISC Estimand Framework in Clinical Trials | Hiral Raval, Simon Luo |
This abstract presents an end-to-end workflow for implementing the CDISC estimand framework in a late phase nephrology trial, with a focus on regulatory submission readiness. The primary challenge addressed is translating complex estimand definitions, particularly those involving intercurrent events (ICEs), into ADaM-compliant analysis datasets and outputs. Our statistical programming team collaborated proactively with statisticians, data standards experts, and study teams to align data collection, analysis plans, and programming early in trial execution. Key activities included defining | intercurrent event, imputation, adam dataset, efficacy summary, composite strategy, while-on-treatment, treatment policy strategy, treatment arm, hypothetical strategy, data standard, regulatory submission, missing value, cdisc | source poster source |
12 |
| 103 | Enhancing Clinical Study Processes Through AI-Driven Documentation Automation | Pinku Hooda | In the evolving field of clinical research, artificial intelligence, machine learning, and large language models are being increasingly utilized to automate the generation and validation of essential documentation. This includes drafting statistical plans, segments of clinical study reports, electronic case report forms, and data standards such as SDTM and ADaM mapping guidelines. Implementing these technologies can significantly streamline the documentation process, reduce manual workload, and improve the accuracy and consistency of clinical data. The use of advanced language models allows for contextually appropriate content creation, aiding researchers in decision-making and documentation quality. These innovations promise to optimize workflows and elevate standards in clinical study management | ecrf, crf, cdisc | source presentation source |
3 |
| 104 | Enhancing Clinical Study Report Authoring with Generative AI: Lessons Learned on TFL Formatting and Metadata Integration in Merck's iRAP Tool | Elisabeth Pyle, Madhusudhan Reddy Papasani |
Merck is leveraging Generative AI to streamline clinical study report (CSR) authoring through the Intelligent Research Authoring Portal (iRAP) CSR tool. This innovative tool accelerates the creation of initial CSR drafts by integrating protocol details, Tables, Listings, and Figures (TLFs), and standardized templates, enabling medical writers to focus on data interpretation and clinical relevance. Pilot studies identified critical requirements for TLF formatting and the associated metadata within Data Meta (DM) files to ensure seamless incorporation of immutable TLFs into both the CSR body and appendix. In response, a comprehensive checklist was developed for Statistical Programmers to follow when delivering TLFs and DM files. This presentation will share detailed insights from the pilot phase and illustrate how adherence to the checklist improves compatibility with the iRAP tool | clinical significance, csr | source presentation source |
4 |
| 105 | Enhancing Clinical Trial Dataset Mapping with AI/ML Techniques: An Approach for CDISC Variable Prediction | Parvee Kumar | In clinical trials, the mapping of datasets (ADAM, SDTM) and CDSIC variables to standardized terminologies, such as those defined by CDISC (Clinical Data Interchange Standards Consortium), is essential but often time-consuming and prone to human error. The application of AI/ML techniques offers a promising solution to automate and optimize this mapping process. Machine learning models, including Random Forest and other advanced algorithms, can be trained to predict domain names and map variable names to the correct standards by identifying patterns in historical clinical trial data. These AI/ML models improve both the speed and accuracy of dataset mapping by analyzing features like variable names, types, and their contextual relationships within clinical trial datasets. By leveraging past clinical trial data, machine learning techniques provide a powerful mechanism for consistent, automated variable mapping, thereby reducing manual intervention and potential errors | clinical trial, machine learning, clinical data, clinical research, proc hpforest, data cleaning, sdtm programming, controlled terminology, regulatory submission, sdtm domain, proc freq, adverse events, cdisc | source presentation source |
7 |
| 106 | Enhancing ISS Analyses with the Continuity-Corrected CMH Method: An SAS-Based Solution for Zero-Event Strata | Ji Qi, Kaixiang Li |
The Mantel-Haenszel (MH) method is a standard approach for estimating treatment effects in stratified analyses and is widely used in Integrated Summary of Safety (ISS) submissions. However, zero-event strata, common scenario in rare adverse, poses challenges, as SAS PROC FREQ does not natively support continuity correction, which potentially leads to biased risk differences and ultimately impact regulatory conclusions. To address this, we developed a custom SAS macro that applies the Haldane-Anscombe correction by adding 0.5 to all cells in any stratum with zero counts, with optional exclusion of double-zero event strata, and output more accurate CMH statistics. Applied to ISS datasets, this approach stabilized estimates of risk differences, reduced bias toward the null, and aligned with regulatory expectations |
odds ratio, relative risk, sample size, confidence interval, sample, treatment effect, adverse events, proc freq, treatment arm | source presentation source |
5 |
| 107 | Enhancing Protection of Confidential Commercial Information through Unified Cross-Functional Strategic Alignment | Sarah Tursi, Nancy Williams |
Presentation | source | 21 | |
| 108 | Ensuring Reliability of Medical Devices Through Statistical Oversight in the Age of AI | Parvee Kumar, Shrutimita Pokhariyal |
As medical devices like smart injectables and wearables become more data -driven, ensuring their reliability is critical-especially in clinical and real-world settings. While Artificial Intelligence (AI) and Machine Learning (ML) offer predictive capabilities to detect device malfunctions or failures early, they must be supported by rigorous statistical frameworks to ensure accuracy and regulatory compliance. This paper highlights the central role of biostatistics in monitoring and maintaining device performance. Techniques such as Survival Analysis (SA) help estimate time-to- failure, while control charts and Statistical Process Control (SPC) methods detect shifts or trends in device metrics like injection force or temperature. Automated data -quality monitoring to prevent missingness and reduce noise (e.g., alerts for late data, out -of-range values, and incomplete records) can further improve the integrity of downstream analyses | data cleaning, survival analysis, data integrity, clinical trial, proc freq, audit trail, missing value, adam dataset, quality assurance, insurance risk score, clinical research, data science | source presentation source |
11 |
| 109 | Equity as a Dimension of Quality: Embedding Equity into Risk-Based Quality Management | Lydia Matombo | Risk-Based Quality Management (RBQM) has reshaped clinical trial oversight by reducing reliance on source data verification, embedding centralized monitoring, and reinforcing quality-by-design principles. While these frameworks have strengthened data integrity and operational efficiency, they have often failed to systematically address equity-related risks such as population representativeness, site accessibility, and health literacy. These gaps contribute to underrepresentation, introducing operational inefficiencies, and weakening the interpretability and generalizability of trial outcomes. This paper presents a practical framework for embedding equity as a measurable, inspection-ready quality risk within RBQM oversight systems. By translating equity considerations into defined risks, key risk indicators (KRIs), quality tolerance limits (QTLs), and proportionate mitigation actions, sponsors and CROs can operationalize equity using existing RBQM tools | inspection readiness, clinical trial | source presentation source |
6 |
| 110 | Estimand-Aligned Multiple Imputation Strategies in CNS Trials with Complex Intercurrent Events | Veera Thota, Ian Lees |
Multiple imputation (MI) is a robust approach for handling missing data in sensitivity analyses, especially when evaluating different intercurrent event (ICE) strategies. In this context, ICE 1 and ICE 2 represent distinct assumptions about post-ICE outcomes, such as treatment discontinuation or rescue medication use. MI enables the generation of plausible values under each ICE scenario, preserving statistical integrity while exploring the impact of varying assumptions. The challenge lies in aligning imputation models with the estimand framework, ensuring that each ICE strategy reflects clinically meaningful interpretations. By implementing tailored imputation models for ICE 1 and ICE 2, analysts can assess the robustness of primary results and quantify unce rtainty due to missingness. This approach supports regulatory expectations for transparency and comprehensive sensitivity analysis, ultimately strengthening the credibility of treatment effect estimates in submissions | imputation, intercurrent event, missing value, multiple imputation, covariance, correlation, treatment effect, while-on-treatment, treatment policy strategy, adverse events | source presentation source |
6 |
| 111 | Evolve Clinical Data Management: A Python-Based Framework for Clinical Data Transformation with Generative AI Integration | Mitul Patel, Denis Sokolov |
data cleaning, clinical programming, clinical data, audit trail, data transformation, data visualization, data governance, clinical data management, proof of concept, dataset, metadata management, response variable, raw data | source presentation source |
5 | |
| 112 | Framework for Incorporating Estimands within the Statistical Analysis Plan | Inderjeet Arora | In clinical trials, estimands establish a structured framework to define the precise treatment effect of interest, ensuring that statistical analyses are aligned with clinical objectives. Within a Statistical Analysis Plan (SAP), the biostatistician's role is to specify each estimand clearly, consistent with ICH E9(R1) guidance. This involves detailing its key components: the target population (e.g., all randomized subjects), the variable or endpoint of interest (e.g., primary efficacy measure), strategies for handling intercurrent events (e.g., treatment policy, hypothetical, or composite), and the chosen summary measure (e.g., mean difference, hazard ratio). The SAP should also outline approaches to missing data and include sensitivity analyses to evaluate robustness. Collaboration with clinical teams is essential to ensure that estimands address clinically meaningful questions | missing value, intercurrent event, sensitivity analysis, treatment effect, imputation, while-on-treatment, data science, treatment policy strategy, multiple imputation, clinical trial, missing not at random, proc means, change from baseline, likelihood function, adverse events, proc mixed, repeated measures, randomization, hypothetical strategy | source presentation source |
9 |
| 113 | From ATC to ACG: Implementing the Concomitant Medication Grouping (ACG) Dataset in Clinical Trial Analysis | Michelle Starkell | concomitant medications, intercurrent event, whodrug, adam dataset, clinical trial, cdisc | source presentation source |
8 | |
| 114 | From Automation to Audit-Readiness-AI's Growing Role in Statistical Programming | Manoj Kumar Maripally | AI-assisted statistical programming is transitioning from a theoretical concept to an operational reality, fundamentally reshaping the delivery of clinical development. By automating repetitive activities such as data cleaning, coding, and validation, AI technologies reduce operational burden and accelerate critical milestones, including SDTM and ADaM dataset development, clinical reporting, database lock, and regulatory submissions. Generative AI, particularly large language models, offers additional value in the early stages of code development and documentation drafting. Nevertheless, robust human oversight remains essential. AI should be positioned as a collaborative tool rather than a fully autonomous solution -enhancing programmer efficiency while preserving the expert judgment required for contextual interpretation, reproducibility, and regulatory compliance | clinical programming, data cleaning, clinical research, machine learning, clinical trial, data mapping, regulatory submission, statistical programmer, cdisc | source presentation source |
4 |
| 115 | From Black Box to Glass Box: A Context-Aware Approach to SDTM Mapping Automation with Interactive Visual Validation | Rostislav Markov, Jacques Lanoue |
We present a specification-first approach to Study Data Tabulation Model (SDTM) automation in which artificial intelligence (AI) assists in generating machine-readable mapping specifications rather than performing "black-box" data transformations without lineage or reproducibility. Our architecture separates data modeling, transformation logic, and specification generation into clearly defined layers. Dataset operations are expressed as transparent SQL-like modeling prior to SDTM mapping and derivation execution. Transformation functions are implemented as versioned Python modules deployed as cloud compute resources and managed through a controlled registry with explicit metadata. Machine-understandable JSON specifications ("MUS") reference only these approved functions and fully describe source datasets, dataset operations, mappings, derivations, and preprocessing steps. The system operates on a reuse-first model | data cleaning, proc sql, sdtm programming, sdtm domain, statistical programmer, cdisc | source presentation source |
12 |
| 116 | From Chaos to Clarity: Can AI/ML Truly Decode Unstructured Clinical Data? | Umamaheswari Adiseshan | Unstructured clinical data, including physician notes, patient-reported outcomes, and especially clinical images, contains high- value information for downstream analysis, biomarker discovery, and long-term research reuse. However, privacy regulations (e.g., HIPAA in the United States and GDPR in the European Union) restrict sharing of sensitive data unless protected health information (PHI) is properly removed. In clinical images, PHI may exist in both pixel overlays (labels, IDs) and metadata (e.g., DICOM headers), creating a challenging de-identification problem at scale. This paper presents a hybrid Generative AI (GenAI) approach combining automated detection/redaction with targeted human review to enable scalable, compliant PHI redaction while preserving clinical utility | feature extraction, data cleaning | source presentation source |
4 |
| 117 | From Complexity to Consistency: Scalable Anonymization Across Clinical Trial Portfolios | Ahmed Eldafrawy, Danielle DiBonaventura |
Presentation | clinical trial, clinical data, life sciences, threshold, options statement, sample | source | 22 |
| 118 | From Data Cutoffs to Decision-Making: Programming Futility Analysis in Oncology Clinical Trials | Mani Mudumba | Interim analyses are increasingly incorporated into oncology clinical trials to improve operational efficiency and support timely decision-making. This paper presents a statistical programming perspective on implementing a randomized Phase 3 study with run in portion and a pre-specified interim futility analysis. We detail key programming strategies for managing interim data cutoffs, deriving futility analysis populations, and supporting both safety and efficacy outputs, particularly when progression-free survival (PFS) is the primary endpoint. The paper explains how the outputs facilitate timely, data-driven decisions by Independent Data Monitoring Committees (IDMCs). Finally, I discuss the programming challenges specific to implementing futility analysis and provide practical solutions to ensure accurate, reliable interim results for decision-making | futility analysis, interim analysis, phase iii, progression-free survival, hazard ratio, clinical trial, data cutoff, kaplan-meier plot, km analysis, clinical programming, randomization, proc phreg | source presentation source |
6 |
| 119 | From Data to Dossier: Lessons from Cross-Company Regulatory Submissions | Nikita Joseph, Gayathri Mahadevan |
Cross-company pooling and submission presents significant challenges requiring agile strategy adaptation, unique problem-solving skills, and predictive insight. Drawing on personal experiences collaborating in cross -company projects this paper will share practical lessons to drive a successful regulatory submission for products co-developed by pharma companies. We will examine complexities around such pooling strategies, and how shifting company - driven priorities impact programming deliverables. We will explore data challenges caused by differing company standards through examples from adverse events, exposure, pharmacokinetics, and immunogenicity data. Particularly, challenges around data integration and alignment, standard adherence, and maintaining analys is consistency will be discussed. We will then shift our focus to programming review of submission documents like pooling strategy, SAP, CSR, and SmPC | sample, adverse events | source presentation source |
7 |
| 120 | From Demo to Deployment: Bridging the AI Implementation Gap | Melanie Hullings, Emily Yates |
clinical research, artificial intelligence, database lock, data cleaning | source presentation source |
5 | |
| 121 | From Git Init to GitHub Push: A Complete Version Control Workflow for Statistical Programming | Jeetender Chauhan | staging area, audit trail, clinical research, clinical programming | source poster source |
14 | |
| 122 | From Legacy to ISS/ISE: Real-World Lessons in Module 5 Submissions | Ajay Gunda, Kishore Pothuri |
Packaging submission components into the eCTD Module 5 (m5) folder structure may seem like a straightforward technical task. It demands a deep understanding of regulatory requirements, therapeutic context, and extensive prior experience. The complexity multiplies when projects include legacy studies, remediated deliverables, and integrated submission packages such as ISS and ISE. Each of these adds unique challenges, ranging from harmonizing metadata across inconsistent datasets to managing updated define.xml files, ADRGs, and cSDRGs, while ensuring compliance with the latest CDISC standards and agency expectations. This presentation will share practical learnings, key challenges, and strategies to efficiently organize SDTM, ADaM, TLFs, and associated submis sion documents within m5. We will highlight common pitfalls, risk mitigation approaches, and best practices to streamline packaging while maintaining submission quality and integrity | adam dataset, clinical trial, cdisc | source presentation source |
5 |
| 123 | From Oversight to Insight - Rethinking Risk-Based Quality Management | Madhumita Ghosh, Eva Kelty |
Risk-Based Quality Management has become a cornerstone of modern clinical trial oversight, but its success depends on how easily teams can access and understand the signals that matter. When critical insights are buried in disconnected listings or locked in isolated systems, identifying emerging risks often relies on experience and manual effort rather than clarity and context. An effective RBQM approach brings data together conceptually, not just operationally. By connecting deviations, safety events, and operational trends within a structure that mirrors how trials actually unfold, teams can explore quality signals dynamically. This enables a shift from chasing after inconsistencies to understanding how different factors relate - supporting earlier interventions and smarter prioritization | clinical trial, protocol deviations | source presentation source |
5 |
| 124 | From Policy to Practice: A Compliance Checklist for EMA Policy 0070 | Anaya Rehman | EMA Policy 0070 mandates the publication of clinical data with the goal of increasing transparency while protecting patient privacy and commercially confidential information. Despite its clear intent, implementation remains complex. Compliance with EMA Policy 0070 remains challenging due to gaps in governance, inconsistent document quality, and weak anonymization and CCI management, often leading to delays and increased regulatory risk. This paper identifies common compliance pitfalls, such as the ownership void, CCI overreach, and inconsistency, and proposes a structured submission readiness or a compliance checklist. This 'playbook' approach serves as a critical tool for organizations without dedicated teams, ensuring regulatory compliance, preserving scientific integrity, and protecting patient privacy and intellectual property | clinical data, execution phase | source presentation source |
6 |
| 125 | From SAS to Cursor: Vibe-Coding into SAS, R & Python | Kevin Lee | The rise of Gen AI is revolutionizing how coding is performed across industry, and "vibe coding" stands at the forefront of this transformation. Coined from Andrej Karpathy's idea of "embracing the vibes" of AI-assisted coding, vibe coding represents a seamless flow between human logic and AI coding agents-where programmers prompt, review, and collaborate with AI to produce code efficiently and intelligently. As an example, Microsoft CEO Satya Nadella estimated in April 2025 that 20% to 30% of Microsoft's code was generated by AI. The presentation explores how vibe coding will reshape programming in Biometrics, where SAS, R, and Python remain essential. It illustrates how conversational AI tools (ChatGPT, Gemini, Claude), AI-native IDEs (Cursor, Windsurf, GitHub Copilot) and customized AI Coding Agents and Agentic workflow enhance traditional workflow - automating coding, debugging, and validation processes while preserving scientific and regulatory rigor | data cleaning, error message | source presentation source |
5 |
| 126 | From SAS to R for TLGs: A Case Study on Implementing an 'R-First, SAS-QC' Workflow for Time-to-Event Analysis | Adler Moisés Fallas Rodríguez | For decades, SAS has been the primary language for generating clinical trial Tables, Listings, and Graphs (TLGs). With the increasing adoption of R in the pharmaceutical industry, our team faced the challenge of transitioning to open-source tools while maintaining compliance and quality standards. We implemented a workflow where R is the first -line tool for TLG generation, with SAS serving as an indep endent benchmark for quality control (QC). This dual-programming approach leveraged R's flexibility and transparency while relying on SAS for validation and regulatory alignment. As an example, we illustrate a Time -to-Event summary TLG. In R, we applied packages supporting metadata -driven table con struction and automated formatting. The same key derivations -patient counts, survival estimates, and hazard ratios -were independently replicated in SAS, providing validation | data cleaning, quality control, hazard ratio, clinical trial, proc phreg, confidence interval, adam dataset, analysis set, independent qc, sas output, data science, proc step, survival analysis, cdisc | source presentation source |
8 |
| 127 | From System Exports to Standards: Will Standardisation of External Vendor Data Redefine Clinical Data Management/Analysis? | Sai Sri Sravya Deevi | In the rapidly evolving field of clinical research, integrating external data sources is crucial for enhancing clinical trial data and gaining insights into patient treatments. Diverse data sources like patient assessments, lab results, genomics, images/videos, and real-world evidence hold significant potential to advance drug development. However, orchestrating these data presents challenges in clinical data management. To successfully integrate external data, it is essential to address challenges in standardization, quality assurance, governance, and project management. Our strategies include adopting CDISC standards, enforcing stringent quality control, establishing clear governance frameworks, and ensuring timely data transfers. This approach allows us to standardize diverse data sources and provide study data for integrated review and analysis in under two months from contract sign-off | data cleaning, clinical trial, controlled terminology, clinical research, quality control, clinical data, cdisc | source presentation source |
5 |
| 128 | From Tables to Text: Revolutionizing CSR Development with Generative AI | Eric Zloty | Clinical Study Reports (CSRs) are essential to regulatory submissions yet remain one of the most time-consuming and manual deliverables. Traditional CSR development involves labor-intensive synthesis of protocols, outputs, and narratives, often leading to inefficiencies and inconsistent quality. This session presents a generative AI framework that combines structured study metadata, statistical outputs, and regulatory templates to dynamically produce draft narratives aligned with ICH E3 guidance. Operating in a compliant GxP environment, the solution delivers consistency, traceability, and reproducibility while dramatically reducing timelines. Attendees will see how AI can interpret data tables, annotate key insights, and generate high-quality drafts for medical writer review. Rather than replacing expertise, AI augments it and freeing teams to focus on strategic interpretation and preparing organizations for the era of digital-first submissions | data cleaning, clinical trial, inspection readiness, audit trail, treatment effect | source presentation source |
10 |
| 129 | Fuzzy Matching DOSE Date Intervals ON Analysis Dates Across SAS (DATA Step, PROC SQL, SAS Macro, and PROC FCMP) | Inka Leprince, Troy Martin Hughes |
In clinical trial analysis data sets, the treatment variable DOSEON, the treatment dose at record start, was introduced in ADaM OCCDS v1.0 for use in non-ADSL ADaM data sets. DOSEON is typically derived from exposure data by determining if and when an analysis (start) date falls within a treatment dose level start-end interval. Although the definition appears simple, performing the merge of exposure data with another analysis data set is complex because the CDISC data structures do not provide a natural primary key for joining these data sets. Instead, the derivation of DOSEON requires matching a single date to a date range. This paper explores procedural and functional approaches to solving this fuzzy matching lookup operation. Procedural approaches require one or more DATA steps and/or SAS procedures, as demonstrated with PROC SQL joins, a SAS macro, and a hash object lookup within a DATA STEP |
hash table, output delivery system, proc sql, adverse events, proc step, lookup, hash iterator, macro definition, proc fcmp, set statement, composite key, dose, order by, proc compare, do loop, clinical trial, proc sort, date variable, key, data merging, treatment period, hash add, missing value, cdisc | source presentation source |
32 |
| 130 | Geocoding with the Google Maps API: Using PROC FCMP To Call User-Defined SAS and Python Functions That Geocode Coordinates into Addresses, Calculate Routes, and More! | Troy Martin Hughes | Software interoperability describes the ability of software systems, components, and languages to communicate effectively with each other, and must be prioritized in today's multilingual and open-source development environments. PROC FCMP, the SAS Function Compiler, enables Python functions to be wrapped in (and called from) SAS user -defined functions (and subroutines), and the full panoply of SAS environments supports FCMP-including SAS Display Manager, SAS Enterprise Guide, SAS Studio, SAS Viya, and the latest Cary show pony, SAS Viya Workbench. Productivity and the pace of development are maximized when existing open-source code, such as Python user-defined functions, can be run natively from a Python interactive development environment (IDE) rather than having to be needlessly recoded into the Base SAS language |
proc fcmp, clinical trial, sas function, proc step, output | source presentation source |
15 |
| 131 | Group, Collapse, Repeat: A Dynamic SAS Macro for Stratification Pooling in Oncology Efficacy Analysis | Venkatesh Burla | Stratified randomization is a common design employed in clinical trials, used to ensure balance across key prognostic factors such as disease risk categories and geographic region. These stratification variables are often carried forward into the efficacy analysis of endpoints like progression-free survival (PFS), overall survival (OS), and objective response rate (ORR). However, sparse events count within certain strata can compromise model stability and interpretability. To address this challenge, we deve loped a dynamic and reusable SAS macro that evaluates events that counts across predefined strata and applies a data- driven pooling strategy when minimum thresholds are not met. The collapsing sequence is customizable and typically proceeds by removing region, combining IMDC (International Metastatic Renal Cell Carcinoma Database Consortium) categories, and collapsing sarcomatoid status | create table, proc sql, efficacy summary, proc step, left join, objective response rate, clinical trial, overall survival, progression-free survival, treatment arm, randomization, adam dataset, %if-%then-%else | source presentation source |
9 |
| 132 | Handle with Care: Study-Specific Assessment of R Package Risk | Shannon Grant | R packages contain statistical goodies but sometimes also pack explosive accuracy and reproducibility challenges. We developed a practical R package risk assessment methodology that evaluates package characteristics and study- specific needs for high-profile statistical computing in a validated R environment at a nonprofit, quasi-academic medical research institution. We begin by reviewing challenges of validating open-source software, and risk management approaches from community and industry leaders. We summarize internal discussions rooted in the R Validation Hub White Paper principles, further refining our scope to specific tasks (randomization lists and protocol endpoint analysis) in individual high-profile studies. The assessment is carried out for each study by statistical leads, applies both institutional and study-specific rules, and is documented in an R usage plan. The plan also describes package-specific risk mitigation and measures to maintain reproducibility | data cleaning, randomization, clinical trial, data science, statistical software, r package | source presentation source |
9 |
| 133 | Harmonising gsm and the pharmaverse: Orchestrating SDTM, ADaM, TFLs and ARS in a Unified R Workflow | Zelos Zhu | Open-source tools for clinical trial reporting have matured within the pharmaverse, offering modular solutions for transforming data into industry standards. In parallel, the gsm suite of packages provides a flexible framework for building pipelines through YAML-driven workflows. We present a gsm- pharmaverse pipeline that integrates sdtm.oak, admiral, CARDS, and gtsummary for SDTM, ADaM, ARS, TFLS, respectively, within a connected system of workflows, each supporting its stage. With YAML control files as human-readable checklists, this framework ensures the right tool is applied at the right time, guiding data from raw capture through analysis datasets to final tables. gsm workflows are both flexible and modular. Workflows can be run by individual contributors (e.g., biostatisticians running scripts) or embedded in automated pipelines (e.g., GitHub Actions or validated Statistical Computing Environments) |
adam dataset, clinical trial, cdisc | source presentation source |
3 |
| 134 | Harnessing AI to Streamline ClinicalTrials.gov PRS Submissions | Anaya Rehman | The Protocol Registration and Results System (PRS) is a cornerstone of clinical trial transparency, yet manual submission processes remain a significant bottleneck for drug developers. This paper explores a transformative approach to ClinicalTrials.gov compliance by splitting the workflow into two AI-augmented pillars: Data Extraction and Structured Data Entry. By leveraging specialized Large Language Models (LLMs) and Human-in-the-Loop validation, study sponsors can reduce administrative burdens by up to 50% while improving data accuracy. We present a roadmap for transitioning from resource- intensive manual transcription to an automated, schema-compliant submission model that ensures regulatory readiness and end-to-end transparency | clinical trial, data cleaning, clinical data | source presentation source |
6 |
| 135 | Harnessing AI: Enhancing Test Data Quality and Efficiency in Clinical Trial Set-Up | Rita Pecuch, Varun Jagadeesh |
Generating and entering test data into electronic data capture (EDC) systems is a critical yet laborious task, essential for validating study designs and numerous downstream systems and programs. We previously leveraged open- source technology to develop a tool to significantly ease the time-intensive effort of manual test data entry and notable efficiency gains have been achieved since its release; however, it falls short in effectively covering the full spectrum of test cases necessary for clinical system set-up and programming. To overcome these challenges, Takeda has sought to build on this innovative software solution to harness the power of large language models (LLMs) to expand the range of covered testing scenarios. During the development process, strategies and best practices for maximizing LLM potential in the clinical test data context have been established, along with valuable lessons learned along the way | edc, clinical trial, error message, data transformation, data validation rules, data cleaning | source poster source |
8 |
| 136 | Harnessing LLM-Augmented Automation for ADaM Dataset Generation | Anil Mogha, Ghourav Garg |
Generating Analysis Data Model (ADaM) datasets remains one of the most resource-intensive and error-prone steps in clinical trial programming. Although CDISC standards and open-source frameworks such as admiral provide structure and best practices, implementation still requires extensive manual coding and deep package expertise. General-purpose Large Language Models (LLMs) show promise for accelerating code development, but without domain context they often produce non-compliant syntax, misuse pharmaverse functions, and generate outputs that are difficult to reproduce or validate. This paper presents a metadata-driven, LLM-augmented pipeline for ADaM dataset generation implemented within the SPARC platform. By embedding ADaM specifications, controlled terminology, and explicit library usage guidelines for packages such as admiral, dplyr, and haven, the pipeline enables in-context learning that guides LLMs toward regulator-ready R code | adam dataset, clinical trial, controlled terminology, data cleaning, cdisc | source presentation source |
9 |
| 137 | Honey, I Shrunk the Workload | Cathal Gallagher | Presentation | excel, error message | source | 18 |
| 138 | How R We Now? Tracking Industry Progress in Adopting R and Open-Source for Clinical Reporting | Andrzej Białaś, Ashley Tarasiewicz |
This paper summarizes results from the 2025 Pharmaverse R Adoption Survey, the second annual snapshot of how sponsor companies use R and open-source tooling across clinical reporting (CRF → submission). This year's survey builds upon our initial assessment conducted nearly 1.5 years ago, allowing for a direct comparison of results and providing insights into how adoption rates, practices, and perceptions have evolved over time. The 2025 edition also adds questions on Python usage in clinical reporting. By presenting these updated findings, we aim to highlight key trends, identify ongoing challenges, and share lessons learned. In addition, we will discuss our continued efforts and experiences in fostering open-source collaborations across the pharmaceutical industry to accelerate innovation and improve clinical reporting practices. The survey also highlights what slows adoption down, primarily resistance to change and package/dependency management | data cleaning, sample | source presentation source |
8 |
| 139 | Hybrid SAS-R-Python Pipelines: Coordinating Multi-Language Workflows | Kala Shivali, Ashwin Chaudhari |
A key challenge for many organizations is keeping different software components coordinated, especially when using a mixed stack of tools. This session introduces a solution that provides orchestration patterns to keep SAS, R, and Python components coordinated. It does this by using shared metadata contracts, ensuring consistent inputs and outputs, and providing harmonized logging for all programs. We show how a centralized package server and role- based access control make it easy for teams to choose the right tool without fragmenting governance. By providing a highly auditable and reproducible framework, this solution helps keep your pipelines compliant and debuggable. Attendees will learn how to unify their data science environments and keep projects moving forward without sacrificing compliance. Keywords: Hybrid pipelines, SAS, R, Python, Workflow orchestration, Reproducibility, Audit readiness | access control, data cleaning, clinical trial, macro definition, role-based access control, audit trail, sas log, inspection readiness, production run | source presentation source |
12 |
| 140 | IMAGING DATA ANONYMIZATION (IN CLINICAL RESEARCH) | Mutiullah Qureshi, Anush Revazyan |
Presentation | structured data, data cleaning, hash remove | source | 24 |
| 141 | Implementing AI in Authoring Plain Language Summaries: Insights from a Live Project Experience | Vidhi Vashisht | Presentation | clinical trial, csr, audit trail, privacy, randomization, data cleaning | source | 18 |
| 142 | In the Room Where it Happens: A Behind-the-Scenes Peek into the Performance Evaluation Process | Toshio Kimura | Statistical programmers are eager to grow their career and advance. They work on more studies and submissions, derive complex datasets, perform advanced statistical analysis, generate beautiful data visualizations and dashboards … and yet, they are skipped over for promotions. They are left wondering as to how they are truly being evaluated. A behind the scenes look at the performance evaluation process will help statistical programmers understand how and who are evaluating their performance and even when these reviews are being performed. There are secret goals beyond my normal work? Define your ABOVE and BEYOND. It's not just my manager evaluating me? Map out your stakeholders. The year end performance review isn't performed at the end of the year? Start with a strong mid-year discussion. Statistical programmers must first understand the performance evaluation process, after which they can devise a strategy to optimize for their success | statistical programmer, service provider | source presentation source |
5 |
| 143 | Incorporating Digital Health Technology (DHT) Data in SDTM - Challenges and Examples | Swarupa Sudini, Sandra VanPelt Nguyen |
As digital health technologies (DHT) like wearables are increasingly used to capture data in clinical trials, sponsors face challenges in integrating and mapping this data to CDISC data standards for regulatory submissions. The diversity of data sources, high granularity, large volumes, use of software algorithms, and inclusion of supportive contextual data complicate this process. Current CDISC guidance offers limited examples, making standardization difficult. Some data may not fit clearly into existing SDTM domains, while other data presents challenges related to how it's collected or reported by devices or vendors. This paper will examine some of the challenges in mapping DHT data, summarize the process used at Pfizer to determine the target mapping, and share examples illustrating these challenges | sdtm domain, vital signs, cdisc | source poster source |
7 |
| 144 | Industry Trends in EU CTIS Transparency Documentation: A Comprehensive Analysis | Kara Groom, Sanjay Bagani |
Presentation | clinical trial | source | 14 |
| 145 | Influence Without Authority: Leadership Lessons for Statistical Programmers | Steven Tan | statistical programmer, clinical programming, data science | source presentation source |
8 | |
| 146 | Integrated Summary of Safety (ISS) Demystified: Practical Considerations for a Successful ISS | Ji Qi, Lily Cheng |
The ISS is a critical component of regulatory submissions to agencies worldwide. It provides a comprehensive view of a product's safety profile across clinical trials. A successful ISS requires strategic planning and unified execution. Key design-phase considerations include defining clear objectives, selecting studies with compatible populations and endpoints, and establishing pooling criteria. Subgroups should be pre-specified, and strategies for managing heterogeneity across study designs must be outlined. In the implementation phase, data harmonization is essential. This includes, but is not limited to, standardizing adverse event and concomitant medication coding, aligning treatment-emergent periods and analysis visit windows, normalizing exposure durations, and unifying reporting units for laboratory or other common tests. Transparency in pooling algorithms and traceability to source datasets are vital for regulatory confidence | adverse events, sample, clinical trial, vital signs, phase iii, population, concomitant medications | source presentation source |
8 |
| 147 | Integrating LLMs Using R Shiny for Clinical Data Review by Ensuring Data Privacy and Validity | Peng Zhang | There is growing interest in open-source analytics and interactive data platforms for healthcare research and clinical applications [1] and R Shiny applications have become the standard for visualizing clinical data. Despite this shift, traditional dashboard interfaces often require significant technical expertise and a deep understanding of data structures, posing challenges for clinical reviewers. To bridge this gap, we introduce DataChat, an innovative R Shiny application that enables conversational interaction with datasets through a natural language interface. By integrating large language models with retrieval-augmented generation (RAG) [2] and leveraging the {ellmer}, {shinychat}, and {ragnar} packages, DataChat allows users to generate dynamic plots and tables while grounding responses in validated statistical outputs | clinical data, data review, privacy, clinical programming, data cleaning | source presentation source |
5 |
| 148 | Interactive and Instant: GenAI-Driven Creation of Clinical Trial Applications | Marcin Dubel | clinical trial, clinical data, data science, data cleaning, pharmaceutical industry, survival analysis, adam dataset, kaplan-meier plot, overall survival, regulatory submission, subject-level data, data pipeline, cdisc | source presentation source |
12 | |
| 149 | Interactive Patient Profiles in R Shiny: Enhancing Narrative Review in Clinical Studies | Danny Hsu, Tzu-Chien Chuang |
Traditional patient profile reports often use static tables with limited interactivity, making it difficult for medical writers and clinical reviewers to explore subject-level data efficiently. To address this, we developed an R Shiny application that enables dynamic, code-free exploration of clinical data from oncology and vaccine studies. The tool integrates CDISC-compliant datasets (DM, AE, CM, MH, FA, EX, LB) and provides interactive visualizations of adverse events, lab data, concomitant medications, medical history, and reactogenicity. Built with R packages-plotly, ggplot2, and DT, it allows filtering by treatment arm and subject ID, with hover-enabled pop-ups for added context. Sidebar panels summarize demographics and diagnosis, while tabbed views organize key safety data. This modular, user-friendly interface improves data clarity, reduces manual effort, and supports consistent, high-quality patient profile generation | adverse events, treatment arm, concomitant medications, medical history, cdisc | source poster source |
4 |
| 150 | Introduction of PharmaForest - Forest of SAS packages | Ryo Nakaya, Hiroki Yamanobe, Yutaka Morioka |
PharmaForest is an open-source repository of SAS packages, which is based on SAS Packages Framework (SPF), that brings together a diverse ecosystem of SAS packages designed to accelerate and standardize programming using SAS. Built to support workflows in clinical development, for instance, CDISC data creation and Reporting TFLs for Clinical Study Report, PharmaForest provides a centralized archive where developers and users can easily access, share, and validate reusable macros and utilities for pharmaceutical clinical development. This initiative not only reduces duplication of effort but also fosters transparency, reproducibility, and innovation across the pharmaceutical programming community. By integrating best practices from individuals and companies, PharmaForest aims to become a trusted hub for collaborative development and aims to establish more vibrant communities | sas packages, dataset, clinical research, pharmaceutical industry, cdisc | source presentation source |
7 |
| 151 | Is Complexity Worth It? A Case Study Comparing Multiple Imputation Strategies | Elnaz Ghadimi, David Bushnell |
Missing data is a pervasive challenge in clinical research and can introduce bias or reduce statistical power when not handled appropriately. Multiple imputation (MI) is widely used to address missingness, yet the practical benefits of more complex MI strategies over simpler approaches remain uncertain. T his study compared seven methods for handling missing data within a repeated measures framework applied to a single clinical dataset, including variations of the Treatment Policy approach, Missing at Random (MAR) and Missing Not at Random (MNAR) assumptions, an observed case analysis, and a two-step MI procedure. Across all approaches, treatment effect estimates were directionally consistent, with no meaningful deviations from the primary analysis. The results suggest that more complex MI procedures do not materially alter substantive conclusions in this context |
imputation, missing value, intercurrent event, multiple imputation, treatment effect, while-on-treatment, treatment arm, sensitivity analysis | source poster source |
4 |
| 152 | Is the AI Coding Assistant Catching Your Vibe? A Practical Assessment of Generative AI Tools in Clinical Data Analytics | Chendi Liao | clinical data, data review, data cleaning, data science, data wrangling | source presentation source |
6 | |
| 153 | Is There Really Any Benefit to Stratified Randomisation in Practice? | Pavankumar Bhagat | RCT's are the gold standard for evaluating interventions in clinical research. Stratified randomization is a common technique used to balance prognostic factors across treatment groups, aiming to increase the precision of treatment effect estimates. However, its effectiveness and efficiency compared to simple randomization with covariate adjustment remain a subject of debate. A simulation study was performed to compare performance of stratified randomization and simple randomization with covariate adjustment under various conditions, including different sample sizes, association levels between covariates and outcomes, and treatment effects. By employing a linear regression model, key performance metrics such as type I error rate, standard error, bias, and coverage probability were assessed. These findings suggest that stratified randomization can offer advantages, particularly in smaller sample sizes. It demonstrated improved statistical power and reduced standard errors | treatment effect, type i error, significance level, sample size, sample, proc calis, clinical trial, randomization, confidence interval, proc reg, treatment arm, regression analysis | source presentation source |
9 |
| 154 | Key Guidelines, Tricks and Experiences for the PMDA and a Comparison with FDA and NMPA CDE Submissions | Ramesh Potluri | Submitting documents to regulatory authorities such as the Pharmaceuticals and Medical Devices Agency (PMDA) and the Food and Drug Administration (FDA) is a complex task requiring careful preparation and an in-depth understanding of regulatory requirements. This paper outlines key guidelines and highlights the differences between submissions to PMDA, FDA, and the Nation Medical products Administration (NMPA) Center for Drug Evaluation (CDE). Additionally, it provides practical tips and shared experiences to equip programmers and regulatory teams with the necessary knowledge for efficient PMDA submission processes and effective handling of regulatory inquiries | data cleaning, adam dataset, data validation rules, clinical trial, cdisc | source presentation source |
5 |
| 155 | Leadership Strategies and Practical Frameworks for Statistical Programming in a Multilingual Era | Shaveta Bansal | clinical programming, data cleaning, exception handling | source presentation source |
8 | |
| 156 | Level Up: Change Management Skills as Your Career Accelerator | Aga Rasinska | Presentation | database lock | source | 13 |
| 157 | Leveraging AI for Clinical Data Transparency: Automating Complex Disclosure Workflows | Woo Song | Presentation | data cleaning, quality assurance, proc freq, adverse events | source | 21 |
| 158 | Leveraging AI for Clinical Data Transparency: Progression of AI Integration in PLS Authoring at BMS | Lauren Haggerty | Presentation | source | 12 | |
| 159 | Leveraging Python for Synthetic External Vendor Data Generation for eCOA and ePRO | Hitesh Raval | External vendor (third-party) data such as ePRO (Electronic Patient Reported Outcomes) and laboratory results etc . plays a critical role in clinical trials but introduces challenges, including data privacy concerns, missing values and non-standardization in Data Transfer Agreements (DTA) and Data Transfer Specifications (DTS). Traditional synthetic data generation oft en relies on real patient data to train machine learning models, posing risks of patient re - identification. This paper presents a novel approach to generating synthetic external vendor data for domains like QS (Questionnaires) and LB (Laboratory Test Results) without using patient-level data. By leveraging DTA/DTS metadata and open-source Python tools, this framework produces realistic, regulation -compliant datasets. It also incorporates synthetic Electronic Data Capture (EDC) forms, such as SV (Schedule of Visits) and DM (Demographics), to mirror real-world data complexity | data cleaning, clinical trial, controlled terminology, machine learning, data integration, clinical data, missing value, correlation, adverse events, subject-level data, clinical research, data validation rules, cdisc | source presentation source |
9 |
| 160 | LLM-Powered Clinical Data Flow with Traceability - From Siloed Pipelines to a Semantic Fabric | Pankaj Attri, Matthew J. Becker |
Clinical trial data submissions for FDA review are often hampered by complex, manual processes. We present a LLM-powered solution s that can help automate the clinical data pipeline and, importantly, establish end -to-end traceability. The solution automates the conversion of raw source data into standardized SDTM, generates analysis -ready ADaM datasets, and creates final TLFs for submission, along with the associated code (used to create datasets and analytical outputs). The other important deliverable of the solution is the lineage graph view. By interpreting the Statistical Analysis Plan, the system attempts to create a direct, visual link between SAP requirements and the datasets that fulfill the requirements. This automates traceability to the study design, providing verifiable proof that protocol and SAP requirements have been met, ensuring compliance and confidence for regulatory submissions | clinical programming, adam dataset, mock shell, cdisc | source presentation source |
8 |
| 161 | Machine Learning Model for Risk Qualification of R Libraries | Felipe Jimenez, Luis Diego Leal Chaves |
Regulatory compliance in clinical trials reporting requires the use of reliable, well-documented, and validated tools. While agencies such as the FDA do not mandate specific software for statistical analysis , organizations must ensure that their tools meet validation standards. To support a more objective and scalable qualification of R open-source libraries, we developed a machine learning framework to assess their risk in the context of clinical reporting. The model classifies libraries as high or low risk based on 14 data quality metrics collected using the riskmetricandtestthislibraries. Data were grouped into threshold, training, and validation sets to build and evaluate models performance. After applying feature selection techniques, three machine learning algorithms were tested. The Support Vector Machine (SVM) model outperformed the others, achieving an F1 score of 0.97 | machine learning, support vector machine, data cleaning, insurance risk score, clinical trial | source presentation source |
13 |
| 162 | Managing Commercially Confidential Information Redaction in Contemporary Regulatory Submissions | Kali Tarascavage | Presentation | source | 14 | |
| 163 | Manual vs AI-Assisted Workflows: Practical Strategies for Complex R Package Development | Vedha Viyash | Presentation | data cleaning, clinical trial, data science, bootstrap, r package | source | 41 |
| 164 | Mapping Minimal Residual Disease (MRD) Biomarker Data in Hematologic Malignancies: Clinical Significance and Data Standards in SDTM and ADaM | Jyothi Nelakurthi | Minimal Residual Disease (MRD) is a critical biomarker in hematologic malignancies, representing residual cancer cells that persist in post-treatment and evade detection by conventional diagnostic methods. Studies indicate that MRD negativity is associated with improved event -free survival (EFS) and overall survival (OS). Mapping MRD data into SDTM domains presents unique challenges due to assay complexity, data variant -specific reporting, and evolving controlled terminology. Domains such as Pharmacogenomics Findings (PF), Genomics Findings (GF), and Laboratory Test Results (LB) are utilized, with SDTMIG 3.4v offering enhanced support for molecular data. In ADaM, MRD data is structured to support longitudinal analysis and survival modeling, with strategies for handling missing data and assay sensitivity thresholds. This paper explores the clinical relevance of MRD, outlines data collection methodologies, and provides practical guidance for mapping MRD data into SDTM and ADaM | sample, clinical trial, overall survival, primary endpoint, sdtm domain, surrogate endpoint, clinical endpoint, cdisc | source presentation source |
8 |
| 165 | Maximising Longevity and Saving Costs Through Refactoring of Statistical Computing Environments | Srinath Viswanathan, Vaithi Bharath |
At Bayer, we have a home-grown Statistical Computing Environment (SCE), built using C & Java on Linux, that has served our needs for the past 20 years. Maintenance for the SCE has become more complex due to outdated technology, retiring engineers, and a shortage of relevant skilled professionals in the market. While evaluating the market for SCE offerings, we have also evaluated the option to refactor the current platform to optimize code, introduce automation, generate LLM- powered documentation, and migrate the platform to the cloud with file systems optimized for supporting advanced analytics while keeping preserving the business functionality and improving the user experience. In this paper we will walk you through our refactoring journey and how this could result in cost savings and better operational efficiency in the near term while we build the enterprise strategy for the longer term with the integration of SAS Viya and Posit components | sas viya, data science | source presentation source |
7 |
| 166 | Metadata-Driven TFL Generation: End-to-End Automation from Protocol to Output | Bill Qubeck | Clinical reporting still suffers from manual handoffs between protocol, specifications, programs, and outputs. This talk presents a metadata-first workflow that starts with structured protocol elements, persists them in an MDR, and drives parameterized TLF templates in the SCE. We describe how analysis definitions, derivations, and codelists propagate automatically into code generation and RTF outputs, with governance gates for review and signoff. We show how lineage visualization supports traceability from a table back to its data and assumptions, and how automated checks detect drift when standards or metadata change. Attendees learn the practical patterns that reduce rework, improve auditability, and accelerate CSR readiness; what metadata is essential, how to structure templates, and which quality controls prevent silent errors | analysis set, controlled terminology, proc means, data cleaning, define-xml, adam dataset, clinical trial, sdtm programming, change from baseline, clinical data, adverse events, cdisc | source presentation source |
12 |
| 167 | Methods of a Fully Automated CONSORT Diagram Macro %CONSORT | Jeffrey P. Meyers | The CONSORT diagram is commonly used in clinical trials to visually display the patient flow through the different phases of the trial and to describe the reasons patients dropped out of the protocol schedule. They are very tedious to make and update as they typically require using different software, such as Microsoft Visio or Microsoft PowerPoint, to manually create, align, and enter the values for each textbox. There have been several previous papers that explained methods of creating a CONSORT diagram through SAS1, but these methods still required a great deal of manual adjustments to align all of the components. The %CONSORT macro removes these manual adjustments and creates a fully automated yet flexible CONSORT diagram completely from data. This presentation is a description of the methods used to create this macro | adverse events, group by, data step, macro call, macro parameter, proc sql, proc freq, macro variable, character variable, unique key, macro loop | source presentation source |
18 |
| 168 | mkheader: An R Package for Automated Generation and Management of Program Headers in Clinical Trial Programming | Laura Frederick | In clinical trial programming, maintaining accurate and up-to-date program headers is essential for regulatory compliance, traceability, and audit readiness. Program headers capture critical metadata such as study name, input datasets, authorship, and revision history. As R gains traction in pharmaceutical programming, the lack of standardized tools for automated header generation within R leads to inefficient, error-prone, manual processes that increase compliance risks. The mkheader R package addresses these challenges by automating the creation and management of program headers tailored specifically for clinical trial programming. It automatically extracts metadata - including study, author details, program date, and R version - and seamlessly integrates user inputs to generate consistent, comprehensive headers | clinical trial, data cleaning, r package | source presentation source |
8 |
| 169 | Modernising Statistical Computing Environments: Scalable, AI-Ready and GxP-Compliant by Design | Amanda Lopuski, Rafael Pereira |
Modernizing Statistical Computing Environments (SCEs) is essential for enabling faster, more flexible, and compliant clinical workflows. This paper presents a case study of how Appsilon partnered with Solid Biosciences to design and implement a validated, cloud-native clinical analysis environment on AWS, leveraging container technology and Kubernetes orchestration. The solution enables secure ingestion, analysis, and reporting of clinical data using open-source tools within a fully GxP-compliant framework. What makes this implementation distinctive is Solid Biosciences' decision to build their clinical development infrastructure from the ground up with R and Python as primary languages, rather than following the traditional SAS-first approach. This paper examines the technical architecture, the advantages of an open-source-first strategy, and the practical lessons learned from building an inspection-ready, modular SCE that accelerates delivery while maintaining regulatory compliance | data cleaning, audit trail, clinical research | source presentation source |
7 |
| 170 | Multi-Dimension Comparison Using SAS in the Context of Counting Hemophilia Bleeding Episodes | Chivia Qianyun Zou, Wenruo Hu |
Hemophilia is a rare genetic bleeding disorder characterized by a deficiency in clotting factors, resulting in prolonged bleeding following injuries or surgical procedures. In clinical research, annualized bleeding rate (ABR) is commonly used primary endpoint to demonstrate clinical benefit in terms of efficacy, which is also acknowledged by health authorities. It is calculated as total number of bleeding episodes divided by duration of analysis period in years. A bleeding episode is defined as any occurrence of hemorrhage that may require administration of BPA or factor. It differs from a bleeding event, which refers to individual occurrences recorded in raw data -- a single episode may comprise multiple events. Per the various combinations of causality, bleeding location, duration, and injection strategy, the counting of bleeding is a multi-dimension comparison. This complexity poses a significant challenge in hemophilia research | clinical trial, temporary dataset | source presentation source |
8 |
| 171 | Multi-Standard Submission - Process, Learnings and Recommendations for Best Practice | Mukesh Garg | data cleaning, raw data, regulatory submission, clinical trial, data pipeline, audit trail, extract, sdtm domain, normalization, clinical data, cdisc | source presentation source |
8 | |
| 172 | Multiple Sclerosis Trials Unpacked: Standards, Endpoints and Programming Experience | Daryna Teteriuk | Multiple Sclerosis (MS) remains one of the most complex therapeutic areas in clinical research, where data standards and statistical accuracy play a crucial role. This article offers a practical roadmap for statistical programmers who are either new to MS studies or looking to deepen their expertise. Readers will discover how the nature of the disease shapes trial design, how CDISC standards and agency guidelines bring consistency and traceability, and which study endpoints are most common in the field. Rather than focusing on theory alone, the author highlights real-world challenges and lessons learned, drawing on more than five years of hands-on programming experience in MS trials. By connecting standards, study design, and reporting, the article provides insights to help programmers bridge raw datasets to meaningful outputs | controlled terminology, clinical trial, regression analysis, adverse events, medical history, proc step, least squares mean, survival analysis, logistic distribution, poisson distribution, proc mixed, cdisc | source presentation source |
12 |
| 173 | Next Generation QC Workflow Optimisation: Current Practices and Future Directions from the Emerging Trends & Innovation Working Group | Matthew Finnemeyer, Caroline Phares |
The QC Process Efficiency Project, part of the PHUSE Emerging Trends and Technologies Working Group, aims to improve Quality Control (QC) process efficiency while ensuring regulatory compliance. Building on previous PHUSE working groups and papers that outlined best practices for QC, this project focuses on not just sharing current practices but also advocating for innovative approaches enabled by new technologies. The first phase of research addresses three distinct topics: global regulatory considerations, similarities and differences in QC practices including risk-based approaches, and recent advancements in technologies and methodologies that are shaping the future of QC. By exploring these areas, the group seeks to pave the way for more efficient, compliant, and technologically driven QC practices. This paper discusses the findings of this research, the challenges identified, and the opportunities for future advancements in QC processes within Data/Statistical Programming | data cleaning, likelihood function, quality control, clinical trial, raw data, clinical programming, data science, artificial intelligence, adam dataset, clinical research, cdisc | source presentation source |
13 |
| 174 | Next-Gen Statistical Programming: AI as Your Coding Co-Pilot in R | Sumesh Kalappurakal, Mark Rothe |
clinical programming, data cleaning | source presentation source |
17 | |
| 175 | Not Another Other: Bridging Data Collection with SDTM Mapping Without Open Text Fields | Aminta Jimenez | Controlled Terminology is essential for ensuring consistency, accuracy, and interoperability in clinical data. The persistent use of free-text fields such as "Other, specify" undermines these goals, introducing ambiguity and complicating downstream mapping to SDTM. This paper advocates the elimination of unnecessary free-text entries in favour of structured data collection aligned with Controlled Terminology. We will demonstrate how existing Electronic Data Capture (EDC) system functionality can be leveraged to guide Investigators toward valid selections that bridge data collection with SDTM mapping. We present a practical methodology for designing case report forms (CRFs) that promote standardized input, demonstrate mapping strategies to SDTM, and explore techniques for representing extensive codelists within Define-XML. We will show how to improve data quality, streamline standardization, and enhance regulatory readiness by minimizing the use of "Other, specify" whenever possible | edc, crf, data cleaning, proc freq, sdtm programming, cdisc | source presentation source |
6 |
| 176 | Objects in Mirror Are Closer Than They Appear - How an Organisation Came Together to Accelerate Open Source Adoption | Eileen Ching, Benjamin Arancibia |
At first glance, the transition to open source in how we manage and analyze data in drug development seemed distant, obstructed by legacy systems and compliance concerns. Yet, as teams united around a shared vision, the organization discovered that the catalysts for change were already at hand. This presentation explores the journey of innovation and transformation, spotlighting the pivotal role of an open source adoption program in fostering cross- functional collaboration, transparent governance, and technical empowerment. Through clear communications, fit-for- purpose tooling, tailored training, responsive support, and policy updates, many impacted by this transition not only were adopters but also became champions of open source values. Real-world examples illustrate how continuous feedback, risk-based strategies, team adaptability, and sustained momentum helped accelerate the journey from aspiration to achievement | regulatory submission, clinical programming | source presentation source |
5 |
| 177 | One eye open, one eye closed: How to approach masking in open-label oncology study | Abhay Kumar | Open-label designs are common in oncology due to the life -threatening nature of the disease. Moreover, ethical considerations make it crucial to avoid delaying or withholding effective treatment, ensuring that patients obtain the best available care. Practical challenges also arise because certain oncology treatments have noticeable side effects or specific modes of action that render blinding impractical. Additionally, open- label trials facilitate closer monitoring of side effects and treatment responses, which is vital in rapidly evolving conditions like cancer, allowing for timely adjustments when adverse effects occur. In order to maintain the integrity of the results and minimize the possibility of sponsor's bias, both real and perceived, a comprehensive data masking may need to be performed to avoid revealing any differences between drug of interest and control arm. This paper outlines different techniques for masking the collected(SDTM) and the analysis data (ADAM) | treatment arm, adverse events, proc step, adam dataset, proc freq, sample, data cleaning, cdisc | source presentation source |
8 |
| 178 | One Workflow to Rule Them All: Uniting Anonymisation and Confidential Information | Nirpal Virdee, Lee Berry |
Presentation | data cleaning, audit trail, clinical trial, inspection readiness, life sciences | source | 14 |
| 179 | Open Source as a Strategic Enabler for EMA Submissions: An R Shiny Approach | Marckenley Mercie | The rise of open-source technologies is redefining how organizations approach regulatory submissions. The European Medicines Agency (EMA) requires Risk Management Plan (RMP) tables as part of marketing authorization applications. Traditionally, these outputs have been developed using SAS in processes that are time-consuming and resource-intensive, limiting efficiency across assets. To address this, we developed an R Shiny application to modernize the creation of RMP tables while fitting within existing workflows. The tool automates table generation, shortens timelines, improves consistency, and strengthens collaboration between programming, statistical, and regulatory teams. Its use across different assets demonstrates how open-source solutions can scale effectively while delivering measurable value. This paper highlights how an open-source solution transformed the delivery of EMA- required RMP tables | data cleaning, adverse events, adam dataset, cdisc | source presentation source |
5 |
| 180 | Open Source Pavilion: Beyond Validation Checkboxes: Building Future-Ready GxP Strategies for Open Source | Michael S Rimler, Eli Miller |
Ever been to a neighborhood block party - where you meet your neighbors, learn about their interests, and potentially find ways to help each other with your own 'around-the-house to do list'? That's what this session is aiming to do. Are you contributing to an industry working group or initiative in the Open Source realm? Are you trying to solve a problem related to, or using, open source technology? This session is intended to introduce members of our community who are doing the same, with the aim to create new connections to either accelerate existing objectives or create new ones | source presentation source |
1 | |
| 181 | Open Source Pavilion: Metadata-driven Automation Interactive | Orla Doyle, Rammprasad Ganapathy |
Presentation | phase ii, data cleaning, sdtm domain, clinical research, group by, system organ class, proc freq, cdisc | source | 28 |
| 182 | Opening Session - Global Data Sharing Platforms: Aligning Frameworks, Pressures and Opportunities | Cathal Gallagher | Presentation | clinical trial, data science, crf, csr, clinical data, randomized controlled trial, data governance | source | 31 |
| 183 | Operationalising ICH E6 R3: A Risk-Based Blueprint for Data Management | Vera Pomerantseva | The release of ICH E6 (R3) represents a fundamental shift in regulatory expectations for clinical trial quality oversight. Risk-based approaches, previously focused primarily on site monitoring, are now expected to extend across all functional review thought the entire lifecycle, including data management, medical review, monitoring, safety review and other. This paper examines the impact of ICH E6 (R3) on expansion of risk-based approaches and outlines a practical framework for enabling an integrated risk-b ased management strategy across planning, execution, and oversight | data governance, clinical trial | source presentation source |
5 |
| 184 | Optimising Flexibility and Consistency in Statistical Programming for Platform Trials: Practical Approaches to Data Cut, SDTM, ADaM and TFL Setup | Kelly Fang, Zhen (Laura) Li |
Platform trials, a type of master protocol trial, provide complexity and adaptability in drug development by supporting multiple sub-studies under a unified infrastructure with sub-study-specific protocols. Though this enhances efficiency and innovation, it poses distinct challenges for statistical programming including dynamic analysis requirements, competing timelines, and adapting to evolving protocols. This poster shares practical, scalable solutions from an oncology platform trial, designed to optimize efficiency, flexibility, and consistency across the data lifecycle. Adaptive strategies for raw data cuts allow rapid, precise data selection for varied deliverables, alongside SDTM specification and programming to harmonize data structures across sub-studies. Our ADaM specification and programming provide a consistent, yet flexible framework across sub-studies for various requirements and separate CSRs; with TLF development streamlining output generation accordingly | interim analysis, raw data, data cleaning, clinical research, resource allocation, quality control, clinical programming, clinical trial, cdisc | source poster source |
6 |
| 185 | Optimising Global Team Dynamics in Statistical Programming | Padma Kishan Korsapati | The pharmaceutical industry is undergoing significant transformation, particularly within the statistical programming function. One major strategic shift among large pharmaceutical companies has been the establishment of globally distributed programming teams. While this model offers clear advantages such as reduced turnaround times, improved quality through continuous handoffs, and cost efficiencies, it also introduces unique challenges. These include communication barriers, cultural differences, time zone coordination, and process alignment. This paper explores the complexities of managing global statistical programming teams and presents practical strategies to address these challenges. By identifying key pain points and offering actionable solutions, the goal is to foster seamless collaboration, enhance productivity, and build a cohesive global programming ecosystem | clinical programming, quality control | source presentation source |
5 |
| 186 | Optimising Output Review: A Targeted and Scalable RTF Validation Framework | Siddharth Kumar | Clinical Programming and Biostatistics teams often work under compressed timelines while delivering submission-ready Tables, Listings, and Figures (TFLs). Independent double programming remains the gold standard for validating analysis datasets; however, the review of final Rich Text Format (RTF) outputs is still largely manual. This becomes particularly challenging during CRO-to-sponsor handoffs, where extensive TFL packages must be reviewed within just a few days. Traditional tertiary QC methods, such as converting RTF files back into SAS datasets and performing PROC COMPARE, demand considerable effort and still fail to identify formatting and presentation-level discrepancies. This paper introduces a scalable RTF Validation Framework that integrates SAS, R, and Python to evaluate RTF outputs directly and detect differences across deliveries |
data cleaning, ods rtf, independent qc, macro variable, proc step | source presentation source |
7 |
| 187 | Optimising R Package Development: A Detailed Guide to Workflow and Best Practices | Tejaskumar Mehta, Rikkin Parekh |
Developing your own R package is a powerful way to systematically organize code, enhance reproducibility, and share analytical tools with a broader audience. This paper will present a detailed, step -by-step guide tailored specifically for beginners aiming to create robust and professional-quality R packages. Readers will learn how to establish a well -structured package framework; design clean and reusable functions and produce thorough documentation that ensures ease of use and clarity. The guide will emphasize the importance of rigorous testing to verify code reliability and will introduce effective strategies for managing package dependencies. Additionally, it will provide practical instructions on how to create your package and hosting it on GitHub, facilitating wider accessibility and collaboration | dataset, sample | source poster source |
11 |
| 188 | Optimising Trial Design Dataset Creation To Minimise P21 Compliance Issues | Rama Kudaravalli | This paper presents key considerations for effectively mapping trial design domains to ensure compliance and reduce issues identified by P21 Enterprise checks. Trial design datasets encapsulate the planned structure of a clinical study, including treatment arms, protocol-defined schedules of subject visits, and other essential design elements. Accurate alignment of these datasets with the study protocol is critical to reflect the intended research objectives and operational plans. Once trial design domains are created, they are validated using P21 Enterprise, and any unresolved issues must be documented in the Clinical Study Data Reviewer's Guide (cSDRG). This paper outlines best practices and strategic measures for mapping trial design domains to ensure high-quality, compliant data and minimize the need for post-submission issue resolution | trial summary, sdtm domain, data validation rules, controlled terminology, data cleaning, clinical trial, cdisc | source presentation source |
6 |
| 189 | Orchestrating Efficiency: Symphony and Composer in Clinical Reporting | Margaret Wishart, Karma Tarap |
Symphony is a next-generation solution that streamlines the creation of mock shells in clinical data workflows by digitizing and automating traditional manual processes. It features a user-friendly Shiny-based interface for building and reviewing mock shells, generating table layouts with underlying R code and structured metadata. Once ADaM datasets are available, users can leverage Symphony to instantly produce submission-ready outputs, reducing redundant programming and speeding up delivery. The innovation is powered by two internally developed R packages: Symphony, which provides the Shiny UI for mock shell creation and review, and Composer, the engine that converts metadata into executable R code for both mock-ups and final tables, listings, and figures (TLFs). Both packages are actively being developed and will be open-sourced once stable, encouraging community collaboration and adoption for more efficient clinical reporting | mock shell, clinical programming, macro definition, data cleaning, cdisc | source presentation source |
6 |
| 190 | Orchestrating Multi-Agent Workflows for Complex Trial Data Integration | Rajesh Hagalwadi | clinical trial, clinical data, data cleaning, data integration, clinical research, life sciences, data wrangling, outlier, data integrity, access control, audit trail, training set | source | 8 | |
| 191 | Outcome Blinding in External Control Arm Construction Using RWD: Challenges and Best Practices | Peishan(Jenny) Han, Lixiao Chen |
propensity score, clinical trial | source presentation source |
7 | |
| 192 | Performance-Driven Risk Oversight for AI Agents in Clinical Research | Laxmiraju Kandikatla | The deployment of AI systems in healthcare demands continuous, risk-aligned oversight to ensure safe and responsible operations. We propose a performance-driven model that calibrates human involvement based on two risk dimensions: metrics risk (including accuracy, precision, recall, F1-score, transparency, bias, robustness, explainability, and usability) and AI system risk (clinical impact and decision consequence). These dimensions are integrated into a 3×3 risk matrix, where the intersection informs the appropriate level of oversight. High-risk Systems require Human-in-Command (HIC) oversight, with final decision authority. Medium-risk systems operate under Human-in-the-Loop (HITL) models, with human supervision and active feedback. Low-risk systems function under Human-on-the-Loop (HOTL) oversight, where humans monitor system outputs and intervene only when anomalies occur | data cleaning, key performance indicator, risk classification, data integrity, type i error | source presentation source |
10 |
| 193 | Perspective on Enterprise AI and Posit Tooling | Philip Bowsher | data science, clinical trial | source presentation source |
22 | |
| 194 | Pharmacokinetic Noncompartmental Analysis (PK NCA) Using R: An Efficient Cost-Effective Alternative to Industry-Standard Software | Jennifer Menda, Haley Evans |
Pharmacokinetic Noncompartmental Analysis (PK NCA) is often performed using extensive, proprietary software with limited flexibility. This poster presents R as a powerful, cost-effective alternative for conducting PK NCA, data manipulation, and visualization. We developed an R script that reads ADPC-like files, calculates PK parameters using packages such as pkNCA1 and pkr2, and generates outputs comparable to conventional tools. Results from multiple studies and administration routes (e.g., extravascular, IV infusion, and IV bolus) confirm consistency and highlight procedural differences. Our workflow enables complete analysis within R, including dataset preparation, parameter calculation, and generation of tables and figures. This approach reduces software switching, streamline processes, and supports reproducibility - making PK NCAs more accessible and efficient for teams of all sizes | pk parameter, nca | source poster source |
5 |
| 195 | Pharmula 1 - Driving Innovation in Reporting | Ellis Hughes | In the rapidly evolving landscape of Pharma, clinical programmers develop key insights in Tables, Listings, and Figures that help teams make decisions about clinical trial results. Taking these important insights to the rest of the team, especially to reuse them, is not trivial. To address this, we developed the BEJEWELED suite of tools, a collection of R packages designed to bridge the gap between our programming teams and non-programming staff, particularly our medical writers. This suite in part comprises of the recently open-sourced sparkle, mirage, and polish packages, each providing essential features to insert and format content seamlessly into Word and PowerPoint documents. In my talk I will share how we created, innovated, and deployed tools to reduce drag, saving valuable time between the creation of outputs and delivery of key reports and presentations | clinical trial, accuracy | source presentation source |
4 |
| 196 | PHUSE Working Groups Highlights | Chris Price | Poster | clinical research, adverse events | source | 1 |
| 197 | Practical Survival Analysis Techniques for Clinical Trials Using SAS | Xue Ding, Tianchen Xu |
In clinical trials with time -to-event endpoints, survival analysis is commonly employed to evaluate outcomes such as progression-free survival (PFS) or overall survival (OS), utilizing different methods under proportional hazards or non-proportional hazards scenarios. This tutorial presents a comprehensive, hands-on guide for biostatisticians and statistical programmers applying survival analysis methods to clinical trial data using SAS . We demonstrate primary analyses using the stratified log -rank test and Cox proportional hazards model, including diagnostic tests for the proportional hazards assumption. For scenarios with non-proportional hazards, we highlight advanced methods such as the Max -Combo test and restricted mean survival time (RMST) | log-rank test, rank test, survival analysis, hazard ratio, proc phreg, restricted mean survival time, macro definition, treatment arm, proportional hazards assumption, model training, treatment effect, overall survival, time-dependent covariate, likelihood function, clinical trial, sample, proc step, randomization | source presentation source |
10 |
| 198 | Pre-Conf as a Data Review Layer: An AbbVie Case Study for Optimised Clinical Data Flow | Gian Prakash, Aman Thukral |
data validation rules, machine learning, clinical trial, clinical data, data cleaning | source presentation source |
2 | |
| 199 | Privacy Considerations of Using RWD from a Statistical Programming Perspective | Joe Xi, Lorraine Fang |
With the increasing use of Real-World Data (RWD) in clinical research, privacy compliance has become a critical focus for pharmaceutical companies. This paper presents practical approaches adopted by the statistical programming team at Bristol Myers Squibb (BMS) to ensure privacy compliance when handling RWD, including company-level policies, data storage and access controls, pseudonymization, and de-identification techniques. The discussion is grounded in recent regulatory guidance and industry best practices | clinical programming, access control | source presentation source |
4 |
| 200 | Privacy-Preserving Methods for Oncology Research Using Differential Privacy | Elena Valkanova | Differential privacy (DP) provides formal, quantifiable guarantees that limit how much any single individual record can influence released statistics or trained models. We benchmarked representative DP definitions (pure DP, approximate (ε,δ)-DP, Rényi DP, and zero-concentrated DP) and common mechanisms (Laplace, Gaussian, and Exponential) on oncology survival settings. Using a real-world right-censored colon cancer dataset formatted in SurvSet and oncology simulated datasets, we evaluated privacy-utility trade-offs under matched privacy budgets (ε∈{0.1, 0.5, 1, 2, 5, 10} with δ scaled to sample size). Utility was assessed via predictive discrimination (Harrell's C-index), overall error (integrated Brier score, IBS), stability of key hazard ratios, and synthetic data fidelity (standardized mean differences and Kaplan- Meier agreement) | sample, hazard ratio, sample size, correlation | source poster source |
12 |
| 201 | Problem-First Agentic AI: Transforming Statistical Computing with Purposeful Automation | Rajesh Hagalwadi | Agentic AI in statistical computing for clinical trials is not about adding complexity but about targeting the real challenges in today's workflows. By adopting a problem-first approach, clinical data teams begin by identifying specific pain points, such as delays in cycle time, compliance pressures, or reproducibility gaps and design agentic AI solutions that directly address them. Using collaborative agents in the cloud, organizations can automate data and statistical programming workflows, maintain traceable results, and provide transparent audit trails. Agile prototyping, iterative improvement, and rigorous evaluation ensure that every automation delivers measurable business value. This abstract provides attendees with a practical framework to judge where agentic AI delivers the greatest impact, where traditional workflows remain sufficient, and how to maintain essential human oversight in regulated analytics | data cleaning, clinical trial, clinical data, clinical research, audit trail, raw data, data science, data integrity, artificial intelligence, privacy, quality control, key performance indicator, clinical programming, data validation rules | source presentation source |
8 |
| 202 | Process and Tools for Supporting Haematology Blinded Patient Disease Response Review | Deepak S. Grewal, Naveen Kommuru |
This paper presents an implementation strategy for process design and building tools for blinded disease response reviews in hematology trials. We translate reviewer requirements into automated, reproducible workflows using SAS, R and develop algorithm specification and functions for diverse data derivations and protocol-specific response assessment windows for various Hematology malignancies. The solution ensures auditability and traceability by keeping a running log of subjects, reported data and highlighting changes from prior reports delivered. This work demonstrates how leveraging various programming tools within team expertise and thoughtful process design can significantly improve the quality, consistency, and transparency of blinded patient data reviews in Hematology trials, ultimately supporting better clinical decision-making and regulatory compliance | ods pdf, change from baseline | source poster source |
11 |
| 203 | Programmatically Modifying a Document Created by ODS Excel: Adding a Password and Concatenating Multiple Worksheets into One | Jeffrey P. Meyers | The Microsoft EXCEL destination was added to the SAS output delivery system (ODS) in version 9.4M3 and offered an improvement over the EXCELXP tag sets destination. The EXCEL destination allows SAS users to create workbooks in the XLSX format instead of the XML format which reduces the file size and unlocks additional functionality. This paper focuses on two advanced tricks when using the EXCEL destination: password protection and clickable navigation. The EXCEL destination can create a "protected" workbook but does not include the option to include a password which allows any user to simply remove the protection from within EXCEL. This paper will show techniques to add password protection to the EXCEL file and how to selectively choose which worksheets or columns are protected. The second focus of this paper, clickable navigation, allows the user to click on a cell within a worksheet to jump to another cell anywhere in the workbook leading to very fluid and dynamic reports |
excel, data cleaning, xls, macro variable, output delivery system, macro loop, dataset, data step, %if-%then-%else | source presentation source |
20 |
| 204 | Prompt Engineering in Gen AI: A Game Changer for the professional development of Statistical Programmers and Data Scientists | Kevin Lee | In the rapidly evolving field of clinical trials, prompt engineering emerges as a game-changer for statistical programmers and data scientists. This presentation explores how professionals can harness Generative AI like ChatGPT to enhance productivity and efficiency. We'll delve into the concept of "Prompt it" replacing "Google it," and examine what prompt engineering entails. The discussion covers techniques such as zero-shot learning, few-shot learning, and chain of thought, demonstrating how to optimize outcomes in content development, information retrieval, code generation, and data analysis. We'll also explore specific use cases in biometrics and address the importance of result validation. By mastering prompt engineering, data scientists can unlock the full potential of AI models, enabling them to navigate the complexities of clinical trials more effectively | clinical trial, data science, statistical programmer, phase iii, sas programmer, data cleaning, adverse events, secondary endpoint | source presentation source |
8 |
| 205 | Prompting for Precision: Leveraging Agentic Workflows to Accelerate Clinical Trial Document Understanding | Jaya Simha Inampudi | Clinical trial success hinges on the thorough comprehension of critical documents like protocols, standards, and Statistical Analysis Plans (SAPs). For statistical and clinical programming teams, mastering this extensive documentation is time-consuming yet essential for daily execution. This paper explores the potential of Large Language Models (LLMs) to accelerate this process. We demonstrate that the key to unlocking this potential lies in sophisticated prompt engineering. By embedding deep contextual knowledge of clinical trial processes into prompts, programmers can query complex documentation with high precision, receiving instant, accurate, and relevant answers. This methodology enables rapid information retrieval, clarifies complex methodologies, and ensures consistent interpretation of standards, resulting in significant efficiency gains and reduced onboarding time | imputation, primary endpoint, adverse events, concomitant medications | source presentation source |
14 |
| 206 | Protecting Commercially Confidential Information (CCI) Under the EU Clinical Trials Regulation | Christa Polidori, Sanjay Bagani |
Presentation | clinical research | source | 4 |
| 207 | QC Without Code: ChatGPT Enterprise for Sponsor Oversight of Outsourced TFLs | Jenny Price, Brad Moore |
adam dataset, mock shell, data cleaning, analysis set, proof of concept, excel, proc contents, proc step, demographics table, independent qc, cdisc | source presentation source |
9 | |
| 208 | R We There Yet? A Readiness-Assessment Framework for Adoption of R and Open-Source Solutions | Narasimhan Kaliyamoorthy | How ready is your organization to use R and open-source solutions for programming and regulatory submissions? Can readiness be assessed systematically and reliably? While enthusiasm and pilot successes are growing, the lack of a standard assessment framework necessitates that organizations define their own workflows, tools, and strategies for adoption. This variability raises critical questions: How can we assure regulators of consistency and compliance? How reproducible are results across environments? Who is accountable for governance? True readiness requires a structured assessment to enable broad, consistent, and confident industry adoption that is sustainable in the long term. This paper proposes a reference framework for readiness assessment across four dimensions: People (talent, training, leadership), Infrastructure (computing environment, security, reproducibility), Process (workflows, validation, quality), and Governance (roles, compliance, change, and risk management) | data cleaning, 21 cfr part 11, pharmaceutical industry | source presentation source |
7 |
| 209 | Ready for Next-Level SDTM and ADaM Compliance with End-to-End Processing? | Sunil Kumar Gupta | From specifications to results, smarter tools enable better SDTM and ADaM compliance for higher quality submissions. This is achieved by anticipating common issues and checklists during the SDTM and ADaM mapping process. While many organizations have SOPs for QC, smarter organizations incorporate continuous monitoring of program development. First, SDTM and ADaM specifications and SAPs are imported for metadata processing and compliance checking. This metadata is essential for cross-referencing the logic of all derived SDTM and ADaM variables. While SDTM and ADaM structures and rules must pass the litmus test, the single story message from raw data must be transparent and consistent with SDTMs and ADaMs. Finally, SDTM and ADaM required variables must be created and all raw data must be mapped to SDTMs. With this end-to-end system, onboarding new members to the programming team becomes easier because the documentation and all related programs can be quickly explained | data cleaning, raw data, independent qc, clinical trial, missing value, controlled terminology, proc freq, adverse events, cdisc | source presentation source |
4 |
| 210 | Real-Time AI-Driven TLF Generation with GAutoTLF | Hao Xu, Tyler Rowsell |
Creating Tables, Listings, and Figures (TLFs) for clinical studies is a common and time-consuming process, particularly due to the requirements of double programming validation. GAutoTLF (Guided Automation of Table Listings and Figures for Clinical Study Reports) is an AI- driven solution that automates this process by integrating generative AI with standard TLF shells, template code and study specific clinical metadata. Users submit analysis requests on standard TLFs through simple prompts and receive executable R code for rapid, consistent TLF generation within minutes. The adaptable and scalable design allows GAutoTLF to be leveraged across various therapeutic areas, providing biometrics teams with substantial improvements in productivity, cost eƯiciency, and quality. This presentation will also feature a brief demo of GAutoTLF's capabilities, showcasing its real-time analysis and transformative impact on clinical reporting workflows | clinical trial, adverse events, clinical data, data cleaning, adam dataset, mock shell, cdisc | source presentation source |
14 |
| 211 | Real-Time Integration of EDC and Third-Party Data and Query Actions within an R Shiny Data Review and Patient Profile Tool | Pavel Burmenko, Terek J. Peterson |
clinical data, audit trail, adverse events, clinical trial | source presentation source |
5 | |
| 212 | Real-World Evidence Pavilion: Interactive Workshop: Bridging Estimands & Target Trial Emulation | Lauren Dang, Matt Baldwin |
Presentation | source | 61 | |
| 213 | Real-World Evidence Pavilion: Interactive Workshop: Bridging Estimands & Target Trial Emulation | Lauren Dang, Matt Baldwin |
Presentation | intercurrent event, adverse events | source | 7 |
| 214 | Real-World Evidence Pavilion: Interactive Workshop: Bridging Estimands & Target Trial Emulation | Lauren Dang, Matt Baldwin |
Presentation | intercurrent event, treatment policy strategy, hazard ratio, adverse events, adam dataset, concomitant medications, hypothetical strategy, cdisc | source | 6 |
| 215 | Redefining Quality: Transforming Internal Reviews into Team Development Tools | Arman Melkonyan | In CRO environments, quality is often equated with compliance and deliverable accuracy. However, true performance lies in how teams operate, communicate, escalate issues, apply domain knowledge in real time. This paper introduces a people-focused framework that redefines internal quality reviews as developmental tools rather than process checks, fostering high-performing programming teams. By evaluating not only code, but also team dynamics, communication, and understanding of project context, internal reviews become opportunities to identify growth areas and implement structured, targeted action plans. Tools such as communication scorecards, code quality rubrics, and learning dashboards provide measurable, actionable insights. A CRO case study illustrates how this approach transformed a struggling team into a reliable, results-driven group capable of consistently delivering high-quality outputs and engaging proactively with sponsors | root cause analysis, adam dataset | source presentation source |
10 |
| 216 | Reduce, Reuse, R Shiny: A Framework for Building Generalised and Extensible R Shiny Applications | Nathan Kosiba | clinical trial, data pipeline | source presentation source |
7 | |
| 217 | Regul-ASIAN Uncovered Navigating Reporting Requirements in Asian Markets | Ben Evans | Presentation | clinical trial, clinical research | source | 42 |
| 218 | Reimagining SDTM Mapping with Agentic AI and Automation | Nathan Johnson | Transforming clinical trial data into SDTM standards remains one of the most time-intensive and manual steps in data preparation. This session introduces a next -gen end-to-end automation approach to streamlining this process with human-in-the-loop supporte d intelligence, including an agentic AI -driven method for determining appropriate transformation logic. In this session, explore the workflow and process for this modern method of mapping automation, including how the framework operates, and the components required of underlying clinical data technologies to facilitate this transformation approach. Join this session to learn how automating specification creation can maintain transparency and oversight while reducing manual effort, accelerating SDTM prepar ation and enabling data teams to focus on higher-value tasks | sdtm programming, clinical data, raw data, cdisc | source presentation source |
4 |
| 219 | Reproducible RWE at Scale: From Fragmented Data to Regulatory-Ready Insights | Agnes Youn | audit trail, clinical trial | source presentation source |
4 | |
| 220 | Reproducible Validation Framework for R & Python in GxP | Bill Qubeck, Kala Shivali |
We present a reproducible framework that pins environments using immutable, containerized images and renv.lock files, which capture package version and hash ID to create a software bill of materials. The system's formal validation process includes executing IQs and OQs, with results and metadata stored for audit. The framework is reviewer- friendly and portable, bundling the entire environment into a self-contained package for regulatory submissions. Controlled upgrades and a robust change management process allow teams to satisfy regulators while preserving developer velocity. This provides a credible and compliant solution for open-source tools in GxP. Keywords: Reproducible research, GxP validation, R and Python, Containerized environments, Software Bill of Materials (SBOM), Installation Qualification (IQ), Operational Qualification (OQ), Open-source analytics in regulated environments, Controlled environment rebuilds, Audit readiness | data cleaning, audit trail, access control, inspection readiness, test set | source presentation source |
10 |
| 221 | Rethinking Data Quality - From Fragmented Checks to Connected Insights | Eva Kelty | Data quality has always been central to clinical research, but today it carries new urgency. With decentralized trials, real-world inputs, and multiple sources feeding complex systems, ensuring accuracy is no longer a back-end task, it shapes every decision that follows. When inconsistencies go unnoticed or context is missing, the impact ripples across timelines, compliance, and ultimately, patient outcomes. Meeting these challenges calls for non-traditional approaches. Instead of relying on fragmented checks and static reports, teams benefit from connected data where relationships are visible and issues surface through exploration rather than escalation. Leveraging innovative technologies such as graph databases and interactive data visualization makes it possible to uncover context, patterns, and dependencies often hidden in traditional structures | data cleaning, clinical research, adverse events | source presentation source |
6 |
| 222 | Revolutionising Clinical Data Transformation: Harnessing GenAI and Open Source for Automated SDTM Conversion | Hao (Harry) Chen, Jian Hua Hang |
The conversion of clinical trial raw data into SDTM format remains a major bottleneck in drug development, often requiring intensive manual effort and delaying the transition from data to insight. This presentation unveils an innovative framework that utilizes Generative AI (GenAI) and open-source technologies in R and Python to fully automate the SDTM conversion process, delivering significant improvements in efficiency, transparency, and CDISC standards compliance. Our solution leverages GenAI for automated metadata interpretation and code generation, and employs open-source libraries to map raw data structures to SDTM domains. Key features include iterative learning that continuously enhances data mapping accuracy through historical mappings, user feedback, and a modular, open architecture enabling seamless integration with existing EDC platforms and leveraging Databricks advanced data analysis features | raw data, clinical data, sdtm domain, data cleaning, clinical trial, data transformation, data science, cdisc | source presentation source |
6 |
| 223 | Risk-Driven AI Oversight in Clinical Research: A Tiered Framework Using ISO, ICH Q9 and FMEA | Laxmiraju Kandikatla, Rajesh Hagalwadi |
Artificial intelligence (AI) is transforming healthcare and clinical research, offering opportunities to enhance efficiency, decision-making, and innovation. However, its safe adoption requires governance practices that adapt to the level of risk. This paper introduces a structured, risk-driven framework for AI oversight. The approach begins with a clear definition of each system's purpose and context, followed by risk assessment aligned with ISO 31000:2018, ISO 23894:2023, and ICH Q9. We employ Failure Modes and Effects Analysis (FMEA) to quantify risks through severity, occurrence, and detection, generating Risk Priority Numbers (RPNs) that inform targeted mitigation strategies. Oversight is then scaled proportionately to the risk and model complexity: Human-in-the-Loop for high-risk applications, Human-on-the-Loop for medium risk, and Human-in-Command for low risk. This framework is both rigorous and pragmatic, preserving human autonomy while strengthening accountability and trust | clinical research, artificial intelligence | source presentation source |
9 |
| 224 | Robust Causal Inference in Real-World Evidence Studies with Double Machine Learning | Yilun Huang, Jing Dai |
Estimating causal treatment effects from Real -world Data (RWD) is frequently challenged by high -dimensional confounding, where traditional methods for adjustment may perform poorly with a large number of covariates and are subject to bias from model misspecification. Double Machine Learning (DML) provides a rigorous framework for estimating causal effects in observational settings. It employs machine learning algorithms to model nuisance functions and uses orthogonalized regression to mitigate bias from regularization errors. It also leverages cross-fitting to produce the final causal estimate that is robust to modeling specifications. This paper introduces the theory of DML and demonstrates its application through a simulated real-world dataset. We illustrate a comparison of DML against other methods, including multiple regression covariate adjustment and propensity score methods | regression analysis, treatment effect, logistic regression, machine learning, logistic distribution, sample, propensity score, sample size, adverse events, multiple regression, monte carlo | source presentation source |
5 |
| 225 | Rule-Based Auto-SDTM Mapping Software in Python, Which Automatically Generates Executable SAS or R Programs | Carrie Zhang, Bill Zhang |
The creation of CDISC SDTM datasets is often tedious and time-consuming, requiring repeated manual programming across studies. Such manual process may lead to inconsistency and non-compliance. To address these challenges, we developed a rule-based SDTM auto-mapping software using Python that leverages a knowledge library of metadata and generic mapping rules between raw datasets and SDTM datasets. The software parses predefined, reusable rules, integrates controlled terminology, and automatically generates executable SAS or R programs to produce final SDTM datasets. In addition, it generates SDTM metadata to facilitate review and regulatory submission. This automated solution significantly improves efficiency, strengthens compliance with CDISC standards, enhances traceability, and reduces both manual effort and human error. The tool supports multiple EDC data sources as well as third-party vendor (TPV) data | sdtm programming, sdtm domain, macro definition, excel, cdisc | source presentation source |
9 |
| 226 | Safeguarding CCI in the Age of AI | Zach Weingarden, James Greene |
Presentation | source | 15 | |
| 227 | Safety at a Glance: Enhancing Participant Monitoring in Oncology Trials with Data Visualisation | Agnieszka Bel | Oncology is among the most complex therapeutic areas in clinical trials. Participants often have compromised health, making them susceptible to adverse events, sensitive to interactions with concomitant medications, and prone to having abnormal assessment results. To ensure patient safety, it is essential to have a system in place that provides comprehensive overview of their data. This enables early detection and accurate evaluation of critical findings, allowing for timely intervention. This presentation will demonstrate how data visualisation dashboards using Tableau can be used to facilitate data and safety review in oncology trials, while also discussing limitations that must be considered. Our approach combines the use of reusable standardised dashboards, including domain-specific views and patient profiles, together with customized dashboards, tailored to trial specific requirements (e.g., different disease evaluation criteria) | adverse events, concomitant medications, data cleaning, vital signs | source presentation source |
10 |
| 228 | Scalable Anonymisation for Complex Clinical Trial Data: A Sponsor-Vendor Perspective | Abby McDonell | Presentation | clinical trial, data cleaning, access control, clinical data, normalization, exception handling, data controls | source | 30 |
| 229 | Scalable R Packages Qualification Using a CI/CD Approach | Natalia Andriychuk, Jon Harmon |
clinical trial, group by, r package | source presentation source |
4 | |
| 230 | Scaling Biometric Teams in Emerging Regions: Lessons from Armenia | Nairi Alikhanyan, Arman Melkonyan |
As the demand for biometric programming grows, sponsors are increasingly looking beyond traditional hubs to build flexible and cost-efficient partnerships. Armenia illustrates how emerging regions can meet this demand, not only in the capital but also in regional cities. Choosing the right location is critical: factors include the availability of highly educated people, strong IT talent, and opportunities to reveal and nurture new professionals. This paper examines how these principles guided the decision to expand into Vanadzor, a regional city with a concentrated IT talent base. A supporting case study shows how the previously presented Self-Reinforcing Loop mentorship model was implemented there within one month, accelerating junior development and sustaining quality. Broader lessons will be shared on how to analy ze regional potential, balance training with delivery, and align local teams with global regulatory expectations to create sustainable, high-performing biometric groups | clinical research, clinical trial | source presentation source |
7 |
| 231 | Scaling for the Future: Transitioning from In-House to Commercial Statistical Computing Environments | Hemu Shere, Majdoub Haloui |
Defining and implementing a Statistical Computing Environment (SCE) in clinical development has long been a challenge. In the absence of comprehensive commercial solutions, many pharmaceutical companies, including ours, developed in-house systems to meet regulatory requirements and support multiple programming languages. Our internal platform has effectively served the business for more than a decade, enabling scalability and an expanded user base. With recent advancements in technology, however, commercial SCEs have become increasingly accessible, offering features that were difficult to achieve with custom solutions. This paper outlines our journey in transitioning from an in-house platform to a vendor-based SCE. We describe the drivers for change, the vendor selection process, and the proof-of-concept evaluation criteria. We also summarize early observations and the phased implementation approach we are currently pursuing | data cleaning, clinical research | source presentation source |
8 |
| 232 | Scaling Interactive Applications for Clinical & RWE Insights Best Practices for Adoption, Compliance, and Long-Term Impact | Chris McSpiritt Domino | raw data, clinical trial, life sciences | source presentation source |
8 | |
| 233 | SCE White Paper 2.0 - A Modernised Framework for the Next Generation of Data Analysis | Mark Bynens | The updated SCE White Paper 2.0 advances the standard for modern analytics platforms in the pharmaceutical industry, emphasizing scalability, technology agnosticism, and comprehensive data integration. Building on previous guidelines this version highlights the importance of a GxP environment that supports multiple programming languages and cutting-edge analytics to accelerate clinical trial reporting and drug approval. It addresses challenges posed by legacy systems and evolving digital technologies, advocating for flexible solutions that leverage cloud, advanced analytics, and real-world evidence. By fostering a multi-faceted eco-system, this update aims to enhance data transparency, regulatory compliance, and analysis efficiency, ultimately transforming traditional clinical data processes into agile, innovation-driven operations | clinical trial, audit trail, clinical research, adam dataset, data cleaning, clinical data, cdisc | source presentation source |
7 |
| 234 | SDTM Automation: Implementation, Rollout and Lessons Learnt | Bruce Chambers, Minda Quiroz |
CDISC standards have traditionally been applied to pharmaceutical trials, but the industry is now recognising their value in Medical Device studies. Alcon aimed to become "standards-ready" for its medical device portfolio while ensuring CDISC compliance for its pharmaceutical programmes. By implementing a unified standards framework and automation tools, Alcon achieved measurable improvements in data quality, and cross-study consistency and timelines. This approach enabled scalable adoption across therapeutic areas and study types, while also reducing manual effort and rework. This presentation will detail Alcon's journey from initial standards development to enterprise-wide rollout, highlighting key challenges, lessons learnt and future ambitions. Attendees will gain practical insights into how automation and governance can transform SDTM production across diverse clinical programmes | clinical data, clinical trial, sdtm domain, cdisc | source presentation source |
6 |
| 235 | SDTM Navigator: An Agent-Based LLM Framework for Automated Specification and Programming | Benzi Alex Mathew | SDTM Navigator, an AI-powered application, automates the intricate process of creating SDTM specifications and programs. This innovative tool leverages a Large Language Model (LLM) with a Retrieval-Augmented Generation (RAG) backend to streamline the path to submission-ready clinical datasets. The process commences by ingesting the study protocol and source data. An initial agent analyzes the protocol to discern trial design domains, creating a foundational study map. Subsequently, this information is channeled to a specification agent that correlates the trial design and source metadata to recommend SDTM domains and generate a draft specification. Throughout this workflow, a critical human-in-the-loop paradigm empowers users to review and modify outputs at each juncture, ensuring digital specifications align with study requirements. Upon receiving the validated output, the final programming agent translates the specifications into executable SAS and R programs | sdtm domain, audit trail, cdisc | source presentation source |
3 |
| 236 | SDTMIG v4.0: Are You Ready For It? | Kristin Kelly | The imminent release of the CDISC SDTMIG v4.0 / SDTM v3.0 represents a pivotal moment for SDTM standards. As a major version update, it introduces significant changes that sponsors should begin preparing for now. Key updates include support for Multiple Subject Instances (MSI), the transition from SUPPQUAL to Non-Standard Variables (NSVs), and metadata restructuring aligned with SDTM Model v3.0. The -BLFL variable has been removed and Sections 1-4, which provide guidance across SDTM domains, have been reorganized for improved clarity and usability. Variables are now organized into Variable Groups, enhancing structure and interpretation. New domains such as Event Adjudication (EA) and Gastrointestinal Findings (GI) further expand the ability to handle more types of data. Also noteworthy are updates to the Protocol Deviations (DV) domain and specimen-based Findings domains | clinical trial, cdisc controlled terminology, protocol deviations, disease response, cdisc | source presentation source |
10 |
| 237 | Semi-automating Case Report Form Annotations Using Python and a Metadata Repository | Alvin Mathew | Accurate annotated case report forms (aCRF) are critical for SDTM development and submissions, yet the existing time-consuming and error-prone manual process is at odds with the needs of the fast-paced study start-up phase. The usual practice requires manually creating individual text boxes, changing fonts and colors, and checking adherence to guidelines, including Metadata Submission Guidelines (MSG), SDTM Implementation Guide, and applicable company standards, across potentially hundreds of pages. To streamline this process, we developed a tool to semi-automate CRF annotation. By leveraging Python, the study's architect loader specification, and our company's internal metadata repository, our tool automatically applies the MSG requirements to annotations filled in a control spreadsheet. This spreadsheet is prepopulated with our standard annotations and allows multiple users to fill in the remaining study-specific annotations concurrently | group by, crf, cdisc | source presentation source |
8 |
| 238 | SpecToSAS: AI-Powered SAS Code Generation from Study Specs | Gargee Vaidya, Lluis Gimenez |
Creating regulatory-compliant ADaM (Analysis Data Model) datasets for clinical trials requires translating complex specifications into SAS programs. These specification files often include hundreds of variables with complex, multi -step derivation rules and dataset-specific logic. Manually translating them is time -consuming, repetitive, prone to inconsistency and error, and can lead to code that is difficult to read and maintain. This increases the risk of misinterpretation and complicates collaborative reviews, diverting statistical programmers from higher-value tasks. To address this challenge, we introduce SpecToSAS, a hybrid AI-assisted system that automates the translation of EXCEL-based specifications into ready-to-review SAS code. By combining deterministic Python automation with a multi-agent AI architecture, SpecToSAS handles both straightforward variable mapping s and complex derivation logic while maintaining transparency and human oversight |
data cleaning, adam dataset, clinical programming, clinical trial, excel, sas programming, statistical programmer, data step, adverse events, group by, sdtm domain, normalization, proc sql, proc step, clinical research, cdisc | source presentation source |
12 |
| 239 | Standardising Exposure-Response Data for Modelling and Simulation Using CDISC Principles and {admiral} | Jeffrey Dickinson | Exposure-Response (ER) modeling is a key tool in assessing the safety and efficacy of new drugs, enabling evaluationoftherelationshipbetweendrugexposure,toxicity,andclinicalbenefit. ERdatasetsoftenresemble those used in Population Pharmacokinetic (PopPK) modeling, sharing features such as numeric covariates, relative time variables, and dependent outcomes. While CDISC released standards for PopPK data in 2023,noequivalentstandardscurrentlyexistforERdata. However,manyofthesameprinciplescouldbe applied. This paper explores ER datasets across domains such as Exposure-Efficacy (EE), with endpoints likeOverallSurvival(OS) andProgression-FreeSurvival(PFS); Exposure-Safety(ES), whichmayinclude specific adverse event frequencies; andTumor Response, which may include measures of tumor size over time | adam dataset, adverse events, data cleaning, cdisc | source presentation source |
12 |
| 240 | Statistical Programming Methods for Evaluating the Impact of Immunogenicity on Safety and Efficacy Outcomes | Dinesh Putluri, Rakesh Padigala |
adverse events, adam dataset, sample, exposure, clinical programming, proc sql, cdisc | source presentation source |
10 | |
| 241 | Straight to the Source: Interactive Visual Dashboards for Data Cleaning | Daniel Smith | Data lies at the heart of any clinical trial, making its accuracy a primary objective. Errors can be costly, consume valuable time for rectification, and compromise integrity. Traditional approaches often rely on manual checks and reviews of derived datase ts, increasing the risk of human error and delaying when reviews can occur. This paper highlights the power of using interactive visualisations based directly on source data to enable earlier and more accurate error detection. Upfront discussions with relevant stakeholders ensure that the visualisations are tailored to each department, allowing domain expertise to drive the process. These graphics are integrated into a centralized dashboard, enabling teams to collaborate in real time. Modern clinical trials have access to a suite of visualisation tools, making the adoption of a visual approach more accessible than ever | proc sql, adverse events, raw data, data cleaning, clinical trial, concomitant medications, data review | source presentation source |
8 |
| 242 | Strategic Approaches to Patient Centricity in Data Science | Sylvia Baedorf Kassis | Presentation | data science | source | 21 |
| 243 | Streamlining Report Reusability in SAS Viya: A REST API Approach | David Weik, Mary Dolegowski |
This paper presents a streamlined process leveraging SAS Viya REST APIs to create a reusable, repeatable method for leveraging existing SAS Visual Analytics reports and applying them to new data sources. To enhance accessibility for users unfamiliar with APIs, we encapsulated the code within a user-friendly Custom Step interface. This Custom Step can be seamlessly integrated into existing Flows to automate data processing and update SAS Visual Analytics reports, or used independently. Additionally, the code can be adapted as a macro, offering flexible implementation options to meet diverse user needs and workflows | sas viya, rest api | source presentation source |
6 |
| 244 | Stronger Together: Tech Pathways to Workforce Resilience in Pharma | Jaimin Patel | The Pharmaceutical industry thrives on discovery and progress through innovation, collaboration, and resilience. While navigating strict regulations, complex data flows, and global teamwork, professionals face sustained cognitive and emotional load. This paper reframes stress management as a strategic enabler of scientific excellence and delivery quality rather than a standalone wellness initiative. We propose an employee-centric, technology-enabled roadmap that integrates digital mindfulness and cognitive behavioral therapy (CBT) applications, wearable heart-rate variability (HRV) biofeedback, virtual-reality (VR) relaxation, and AI assistants that reduce administrative burden and rework. In parallel, policy-tech approaches-such as right-to-disconnect norms, meeting hygiene, and capacity-aware work design-embed sustainable balance into everyday operations | inspection readiness, data cleaning | source presentation source |
11 |
| 245 | Submitting eCRT SDTM Data for ISS | Aida Swierk | The creation of electronic Case Report Tabulations (eCRT) for an Integrated SDTM (iSDTM) submission is a complex, cross-functional governance and decision ownership layered process that requires rigorous planning, coordination, and metadata governance across multiple clinical studies. Unlike single-study submissions, Integrated Summary of Safety (ISS) eCRT development demands early alignment on study scope, harmonization strategies, controlled terminology, and the standards that will govern the pooled submission. Differences among contributing studies in CRFs, SDTM specifications, derivation logic, and controlled terminology present significant challenges that must be resolved to ensure a consistent, regulatory-ready package. A central component of ISS eCRT development is the define.xml, which must ensure traceability by accurately representing pooled origins, integration-level assigned values, derivations, remapped values, and value-level metadata | controlled terminology, metadata management, data cleaning, cdisc | source presentation source |
10 |
| 246 | Submitting RWD: Where We Are and Where We Are Going | Ingeborg Holt, Sarah Ferko, Jeffrey M. Abolafia |
The rapid increase in the use of non-interventional study data from real world sources to support marketing applications has posed numerous challenges for both regulatory reviewers and sponsors. This presentation examines the current regulatory environment and proposed standards for submitting Real World Data (RWD). We review responses to the FDA's request for comments in the Federal Register Notice (FRN), Docket No. FDA-2025-N-0287 titled "Exploration of Health Level Seven Fast Healthcare Interoperability Resources (FHIR) for Use in Study Data Created From Real-World Data Sources for Submission to the Food and Drug Administration" to evaluate the expected impact of two proposed options for submitting RWD, expanding CDISC standards to accommodate evidence derived from RWD, and submitting RWD using the HL7 FHIR standard. We also provide short term and long recommendations for submitting randomized controlled trials and RWD study data | csv, clinical research, clinical trial, clinical data, data cleaning, data standard, pharmaceutical industry, regulatory submission, missing value, adam dataset, adverse events, normalization, cdisc | source presentation source |
24 |
| 247 | Supporting Clinical Trial Recruitment Using Real-World Data | Yi-Chien Lee | Clinical trial recruitment is a resource-intensive process and can be even more challenging if no suitable subjects are identified after the site opens. Assessing feasibility before the study starts is crucial as it helps the project team design appropriate inclusion and exclusion criteria and plan the project timeline based on the estimated recruitment speed. In this poster, we use a recent project to showcase how to utilize real-world data to understand the target population in the US. Age and gender information is used to understand the target demographic. ICD-9/10 diagnosis codes are used to evaluate the medical history; NDC/HCPCS claims are used to capture treatment use; and LOINC code/test names/results are used to understand the patient's lab findings. By exploring the potential subject pool size using various inclusion and exclusion criteria, researchers can identify the optimal balance-targeting the appropriate population without being overly restrictive | sample size, clinical trial, sample | source presentation source |
2 |
| 248 | Teaching Small Models To Think Big: Enterprise-Ready AI Through Knowledge Distillation | Ramanathan Srinivasan, Angu Sundaresh Krishna Kumar |
probability distribution, clinical trial, phase i | source presentation source |
7 | |
| 249 | The AI Reckoning: Why Statistical Programmers Must Evolve Now | Bhavin Busa | SAS and R programming are not going away; they will remain foundational to Statistical Programming. Yet the function is rapidly evolving as Artificial Intelligence (AI) and commercially off -the-shelf (COTS) tools reshape daily workflows. The future demands programmers who can integrate Large Language Models (LLMs) into standard-driven solutions, ensuring compliance, consistency, and scalability. CDISC standards play a pivotal role in this transformation. Models such as the Unified Study Design Model (USDM) and the Analysis Results Standard (ARS) are enabling AI-driven applications that automate mapping specifications for SDTM/ADaM and accelerate TFL development. LLMs will further advance this shift by supporting metaprogramming and assisting in the generation of SAS and R code seamlessly integrated into existing workflows | clinical programming, statistical programmer, data cleaning, cdisc | source presentation source |
6 |
| 250 | The Biopharmaceutical Industry Leadership's Guide to Post-Acquisition: Dos and Don'ts | Shereen Anisa Khwajazada | If you are new to this industry or well seasoned, acquisitions are common practice. This guide is not just a collection of theoretical principles, but a reflection of my personal journey in the biopharmaceutical industry. With the benefit of hindsight, I'd like to share practical leadership lessons after an acquisition, drawing on my own experiences. Clear dos and don'ts, as well as the industry-specific integrations that serve as the key to success. Based on personal real- world experience, leaders may find that following these principles inspires a smoother transition and facilitates a successful integration following an acquisition. This guide may also serve as a survival resource for professionals at all levels | data cleaning, ich, inspection readiness | source presentation source |
6 |
| 251 | The Clinical Data Lake: Empowering the Data Caterer Vision with Veeva, SAS and Open Source | Sascha Ahrweiler, Holger Dach |
The Data Caterer vision transforms programmers into curators who prepare and deliver actionable insights. At Bayer, our clinical data lake makes this shift possible. Veeva acts as the structured pantry, ensuring organized, traceable data; SAS provides reliable analytics for regulatory-grade outputs and automated analyses; * Open Source tools bring flexibility and innovation, supporting machine learning and modern visualization. Early implementation highlights cover data flows, compliance safeguards, and readiness for AI. By blending trusted technologies with new approaches, we allow teams to focus on delivering high-quality insights rather than debating tools. This technical deep dive explores how these foundations work together, sharing practical lessons from Bayer's journey. The result: an environment where clinical data is not just stored but curated and served to power better decisions, embodying the Data Caterer vision with Veeva, SAS, and open source solutions | data lake, clinical data, statistical programmer, sas viya, clinical research, data cleaning, data science | source presentation source |
7 |
| 252 | The Curious Case of External Controlled Arms (ECAs): Practical Solutions for External and RWD Integration | Manish Deole, Preethi Samala |
The use of External Control Arms (ECA) in clinical trials is increasing, particularly for rare diseases where typical Randomized C ontrolled Trials (RCTs) may be difficult. Recent FDA guidance emphasizes both the potential and challenges of ECAs, emphasizing data reliability, bias mitigation, adherence to CDISC SDTM and ADaM, statistical approaches such as propensity score matching, and regulatory communication. Additionally, CDISC and PHUSE have released guidance on integrating Real -World Data (RWD) into CDISC datasets. Here, we will summarize key insights from these documents regarding the use of RWD for ECAs and showcase two case studies on integrating ECA data into CDISC-compliant datasets: (1) constructing an ECA using natural history studies and past RCTs and (2) leveraging publicly available RWD and RCT sources, such as the Critical -path Alzheimer's Disease database | propensity score matching, propensity score, clinical research, phase iii, cdisc | source presentation source |
6 |
| 253 | The EQ Matrix: Unlocking the Human Edge in an AI World | Pratibha Jalui | artificial intelligence, life sciences | source presentation source |
4 | |
| 254 | The Journey Beyond Anonymisation: Exploring the Secondary Use of Clinical Trial Data | Yuliia Cherednichenko, Tetiana Skorokhod |
Having anonymized data from over 200 clinical trials, we have gained substantial experience and a deep understanding of the challenges, best practices, and actual strategies required to ensure data privacy while maintaining data utility. However, an important question has consistently guided our curiosity: how is anonymized clinical data actually used once it has been shared or made available for secondary purposes? This paper reviews recent research on the secondary use of clinical data, focusing on its benefits, challenges, ethical concerns, and technical aspects such as data integration, interoperability, and reuse of both structured and unstructured data. Also, we would like to investigate how secondary data use can enhance real-world evidence generation and inform clinical decision-making. Through practical examples and current research trends, we will explore where and how secondary clinical data is being utilized and what the envisioned future progress is in this field | clinical trial, clinical data, clinical research, data protection, data cleaning, data wrangling, data integration, data governance, data standard, imputation | source presentation source |
18 |
| 255 | The Journey of Deriving the Best Overall Response (BOR) for Solid Tumours - A Programmer's Perspective | Milan Adesara, Vishal Gandhi |
The Best Overall Response (BOR) is a critical endpoint in clinical trials for solid tumors, serving as an important measure of treatment efficacy. However, the determination of BOR can be complex while using different response evaluation criteria, such as RECIST 1.1, mRECIST 1.1, and iRECIST 1.1. Each criteria has its own methodology for assessing tumor response, which can lead to variation in how BOR is derived. Given its significance in trial outcomes, it is essential to understand the variation of BOR derivation using different criteria. Moreover, an often-overlooked factor in BOR derivation is the patient's trial completion status. This status can provide additional information especially when the response confirmation is required in subsequent assessment | clinical trial, objective response rate | source poster source |
7 |
| 256 | The Revolution of Digitised Study Designs: Standards, Use Cases and Tools | Berber Snoeijer | clinical trial, data standard | source presentation source |
3 | |
| 257 | The SMART + Framework for AI Systems | Laxmiraju Kandikatla | Artificial Intelligence (AI) systems are now integral to clinical research processes, such as automated adverse event detection in clinical trials, patient eligibility screening for protocol enrollment, and data quality validation. While these technologies enhance operational efficiency, they introduce new challenges regarding safety, accountability, and regulatory compliance. To address these concerns, we introduce the SMART+ Framework, a structured model built on the pillars of Safe, Monitored, Accountable, Reliable, Transparent, and augmented with Privacy, Data Governance, and Guardrails. SMART+ offers a practical, comprehensive approach to evaluating and governing AI systems within regulated environments. This framework aligns with GxP, ICH, and evolving regulatory guidance by integrating operational safeguards, oversight mechanisms, and strengthened privacy and governance controls | data cleaning, clinical research, data governance, data drift, access control | source poster source |
9 |
| 258 | The Trials of Trial Discovery Is Your Registration Hurting Recruitment | Francine Lane | Presentation | source | 17 | |
| 259 | The Use of Git in Statistical Programming | Rita Pecuch, Jim Coates |
clinical programming, audit trail, statistical programmer, data cleaning, clinical trial, protocol amendment, pharmaceutical industry | source presentation source |
9 | |
| 260 | The Year of Yes! How saying "Yes" Ignited GSK's R Adoption | Becca Krouse, Benjamin Arancibia |
clinical programming, regulatory submission | source presentation source |
5 | |
| 261 | Thriving in Change: Building People, Process and Technology Capabilities in Statistical Programming | Ajay Aiyer | The pharmaceutical industry is rapidly evolving, driven by new business models, regulatory changes, and technological advancements. For statistical programmers, success in this environment depends on three core capabilities: people, process, and technology. People capability comes from building resilience and an adaptive mindset - enabling teams to navigate uncertainty, foster psychological safety, and continuously develop skills to meet emerging needs. Process capability requires embracing innovation and adoption strategically, using structured change management, stakeholder engagement, and low -risk pilots to integrate new tools and workflows effectively. Technology capability involves leveraging AI and automation to enhance efficiency, streamline coding and documentation, and manage growing data complexity - while maintaining human oversight for accuracy and compliance | clinical programming, clinical trial, clinical data, data integrity, data cleaning | source presentation source |
5 |
| 262 | Time to Roclet! admiral's use of custom roclets to enhance and extend examples in function documentation | Benjamin Straub | R package function documentation is singular in open source. Detailed argument definitions coupled with inline examples can help breakdown complicated functions that help solve problems. In admiral, R package for ADaMs, we have always prided ourselves on top-notch function documentation. However, as the examples grew for each function the documentation became unwieldy. Using the roxygen2 package, a key component of any R package, we were able to extend the roclets to better display and curate the examples for each function. A roclet is a specific parser used by the roxygen2 package in R to extract documentation from special comments within R code, automatically generating .Rd help files and other documentation components for an R package. This presentation/paper will discuss the technical details of how we extended the roclet and how this positively impacted our function examples | adam dataset, imputation, cdisc | source presentation source |
9 |
| 263 | Transforming Clinical Trials: Present Innovations and Future Paths with AI and Agentic Tools | Sandeep Purwar | clinical research, data cleaning, clinical trial, clinical programming, data governance | source presentation source |
5 | |
| 264 | Transforming Legacy Data into CDISC Standards: Practical Approaches and Regulatory Insights | Khalif Halani, Veena Palle |
Converting legacy clinical trial data into CDISC formats is essential for regulatory compliance and data reuse. Legacy datasets, often captured in older, study-specific formats such as EXCEL, CSV, txt, or SAS, lack the standardization required by regulatory agencies. In a typical setting, SDTM and ADaM datasets serve as the base for the tables, listings, and figures (TLFs) included in the clinical study report (CSR). In some cases, the TLFs for the CSR are not written from CDISC-compliant data due to a variety of reasons. However, there may be a need for CDISC-compliant data after the TLFs are finalized. In this presentation, we will explore practical strategies for converting legacy clinical trial data, where the TLFs for the CSR were not based on SDTM and/or ADaM, into CDISC-compliant formats. We will highlight common challenges and share key regulatory perspectives from agencies such as the FDA and PMDA |
raw data, adam dataset, clinical trial, data science, cdisc | source presentation source |
6 |
| 265 | Transparent Models or Black Box? Using Generative AI to Generate Real-World Intelligence | Sherrine Eid | The rapid adoption of large language models (LLMs) and machine learning (ML) in pharmaceutical research presents opportunities for accelerating clinical development, pharmacovigilance, and regulatory submissions, while underscoring the need for robust model governance. Using synthetic patient data (e.g., Synthea), descriptive summary statistics and automated analyses can be generated across Python, R, SAS, and Lua environments to ensure reproducibility and transparency. Governance frameworks must emphasize model documentation (e.g., model CARDS), data lineage, explainability, and validation under Good Machine Learning Practices (GMLP) and GxP requirements. In current compute environments, containerized MLOps pipelines with embedded audit trails, risk-based monitoring, and human-in-the-loop oversight enable regulatory-grade deployment. Practical approaches include bias testing, drift detection, and controlled retraining aligned with ICH quality guidelines |
data cleaning, machine learning | source presentation source |
7 |
| 266 | Trends in eSub Evaluation Findings: A Twenty-Year Perspective on Data Traceability | Lisa Kaye Brooks | Iris has evaluated hundreds of study data eSubmission packages over the last 20 years. These reviews utilized the publicly available validation software available at the time. There was a focus on checking the qualitative aspects of a compliant submission. These checks were informed by standards, guidance, industry publications, interactions with Food and Drug Administration (FDA) and expert debate in community forums. This paper will share trends in our findings over the years and provide insight into persistent issues that Biometrics Groups and Clinical Research Organizations are struggling to get right in the draft deliverables. We observed a consistent downward trend in traceability-related error rates over time, particularly in ADaM packages following FDA enforcement in 2021; however, persistent errors in origin and derivation documentation remain. This analysis has shown that those that create define.xml for submission would benefit from training as well as improved QC processes | data cleaning, data lineage, outlier, clinical trial, clinical data, quality control, clinical research, excel, cdisc | source presentation source |
9 |
| 267 | Turning Trial Data into Patient-Centric Insights | Catharina Dahlbo, Eva Kelty |
Every clinical data point represents more than a number; it represents a patient's experience. Yet, traditional tools often reduce this experience to static tables and slow processes, making it harder to act when it matters most. This session will highlight how a connected, intuitive approach to exploring clinical trial data can support teams in working more efficiently and uncovering insights precisely when they're needed. Through intuitive visualizations and concept-driven navigation, users can quickly detect safety signals, monitor trends across domains, and explore relationships such as adverse events, lab results, and patient-reported outcomes - all in real time and without coding barriers. By empowering teams to answer questions faster and respond proactively, this approach puts participant well-being at the heart of decision-making | adverse events, clinical trial, clinical data | source presentation source |
5 |
| 268 | Unseen, Unstoppable: The Art of Invisible Leadership | Johnson Dsouza, Syed Gayaz Basha |
Background: Leadership within clinical trial programming and data sciences is often equated with formal authority and managerial roles. Leadership within clinical trial programming and data sciences has traditionally been associated with formal titles, hierarchical authority, and visible decision-making roles. However, many of the most impactful contributors operate without positional power, influencing outcomes quietly through trust, expertise, and collaboration. This paper introduces the concept of Invisible Leadership, a practical, experience-driven leadership model that enables professionals to lead from any role, influence across functions, and drive measurable impact without formal authority. Drawing from real-world experiences in clinical programming, statistical operations, and cross-functional delivery teams, this paper presents an Invisible Leadership Framework and a 4-part Influence Toolkit | clinical programming, clinical trial | source presentation source |
4 |
| 269 | Utilizing AI for Automated Conversion of ADaM Specifications to R Code | Alice Ehmann, Kjersten Offenbecker |
The advent of artificial intelligence (AI) in clinical data sciences is paving the way for enhanced efficiency and accuracy in statistical programming. This paper delves into the application of AI technologies in automating the conversion of ADaM (Analysis Data Model) specifications into R code, a vital process in preparing analysis-ready datasets in clinical trials. Our approach leverages AI models, equipped with natural language processing (NLP) and machine learning capabilities to interpret detailed ADaM specifications and semi-automate their conversion into R scripts. This not only accelerates the development of analysis datasets but also fosters greater consistency across projects. A key feature of this system is its interactive, iterative process that involves data professionals working alongside AI-generated code | adam dataset, clinical programming, quality control, regulatory submission, data cleaning, cdisc | source presentation source |
9 |
| 270 | Validating AI in GxP: Test Harnesses, Guardrails and Compliance Controls | Bill Qubeck | AI adoption fails without credible validation. This article provides a fit-for-purpose validation approach for LLM-assisted authoring and review tools: requirements decomposition, risk assessment, testable acceptance criteria, and a repeatable test harness with seeded scenarios. We capture prompts, inputs, outputs, and adjudications as validation evidence and monitor model behavior over time to detect drift. The design integrates model CARDS, usage policies, and access controls so responsibilities are clear. Attendees learn how to convince QA that AI sits inside a controlled process, which risks must be mitigated, and how to keep validation lean enough to evolve with rapid AI change while meeting inspection expectations. INTRODUCTION: THE REAL PROBLEM WITH AI IN CLINICAL TRIAL SUBMISSIONS AI adoption is not failing in clinical trial regulatory submissions because it is technically infeasible |
data cleaning, clinical trial, csr, data integrity, adam dataset, audit trail, cdisc | source presentation source |
14 |
| 271 | Validation Strategies for Agentic AI Platforms in GxP-Regulated Clinical Data Environments | Rajesh Hagalwadi | Agentic AI platforms, composed of collaborating digital agents, are increasingly used to orchestrate complex clinical data workflows ranging from data ingestion and transformation to quality checks and documentation. In GxP-regulated environments, these capabilities must be validated with the same rigor as other computerized systems, while addressing new challenges introduced by learning components, multi-agent interactions, and continuous change. This paper presents a problem-first, risk-based framework for validating agentic AI platforms used in clinical data programming and statistical computing. Since late 2025, regulators and industry bodies have sharpened their expectations for AI in drug development, emphasizing intended use and context of use, risk-based credibility across the lifecycle, human-AI interaction, and transparent information about performance and limitations | data cleaning, clinical data, audit trail, data integrity, 21 cfr part 11, outlier, data ingestion, clinical research, key performance indicator | source presentation source |
9 |
| 272 | Visualising Essential Metrics for Assessing Disease Dynamics in Hematologic Oncology Trials: A SAS Graphic Approach | Jiashan Li | Chronic Lymphocytic Leukemia (CLL) and Small Lymphocytic Lymphoma (SLL) are different manifestations of the same disease. Patients with CLL and SLL can exhibit lymph node involvement, leading to lymphadenopathy due to the accumulation of abnormal lymphocytes. Absolute Lymphocyte Count (ALC) and Sum of Products of Diameters (SPD) for target lesions are both important metrics for assessing disease dynamics in CLL and SLL. These measures are essential for indicating disease progression, evaluating treatment responses, and guiding clinical decision-making. To enhance the understanding of these data, trends in ALC and SPD over time are monitored using dual-axis plots. This paper presents a comprehensive approach utilizing SAS graphics to illustrate the trends of ALC and SPD, detailing the processes of data collection and data visualization. Ultimately, this work aims to contribute to improved insights in hematologic oncology trials | scatter plot, proc sgplot, graph template language, series plot, adam dataset, cdisc | source presentation source |
8 |
| 273 | Voice-Driven Agent-Assisted Open-Source Data Science | Philip Bowsher | bar chart, adverse events, clinical research | source presentation source |
10 | |
| 274 | What Patients are Interested in when it comes to Data Sharing | Deborah Collyar | Presentation | clinical trial, adverse events | source | 18 |
| 275 | What Pharma Can Learn from Data Science Outside the Industry | Kris Wenzel, Liz Cole |
gold layer, data cleaning, adverse events, regulatory submission, silver layer, data science, bronze layer, clinical data, edc, adam dataset, raw data, cdisc | source presentation source |
11 | |
| 276 | Where are my Agents? Automating Cohort Building with a Dynamic Cohort Builder | Mary Dolegowski, Sherrine Eid |
Defining and building patient cohorts involves a mix of automated and manual methods that leverage structured data, unstructured data, and advanced analytics. Producing reliable real-world evidence that informs clinical practice, supports regulatory decisions, and ultimately improves patient outcomes is imperative for regulatory acceptance of integrated evidence packages. Precisely defining patient cohorts affects study validity and reliability by reducing bias and confounding, ensuring that the evidence generated is robust and valid. This enables meaningful comparisons between treatments and helps in drawing reliable conclusions about effectiveness and safety. Regulators and payers increasingly rely on RWE to complement randomized clinical trials (RCTs) in decision-making. There is a plethora of examples illustrating how well-defined cohorts support regulatory submissions, label expansions, and post-marketing surveillance | sas studio, inner join, create table, macro variable, unstructured data | source presentation source |
12 |
| 277 | You Can't Just Cut the Gordian Knot: Approaches to Real-World Data Quality | S. Robert Collins | Data quality issues prevent a broader use of real-world data (RWD). We begin with an examination of factors that impact RWD in the health care setting. Next, we discuss how methods for extracting and transmitting data can impact data quality. Finally, we look at processes that preserve data fidelity while integrating sources and creating analytic data. Privacy issues complicate matters even further with varying standards between locations. We present an approach to evaluating and addressing RWD quality that begins with more traditional approaches and then look at ways to apply AI to these processes and examine the potential to use synthetic data to enhance privacy while also addressing data issues and bias. In the process, we will consider the challenges and benefits of integrating proprietary and open-source tools to address these issues while ensuring governance, transparency and compliance to standards while aligning with the goals of other RWD efforts | missing value, data cleaning, record count, txt, proc sql | source presentation source |
4 |
| 278 | Your Brain on AI: What We Gain, What We Risk, and How To Harness AI Without Losing Trial Integrity | Aditya Gadiko | clinical research, clinical trial, data cleaning | source presentation source |
6 | |
| 279 | Your Next Teammate Has Arrived: Our Story & Reflections on Building AI Coding Assistant for Clinical Insights Generation | Vincent Shen | data science, clinical data | source presentation source |
4 | |
| 280 | {gsm} Operations - Implementing an Open Central Monitoring Framework | Jenn Krohn | The Good Statistical Monitoring ({gsm}) framework, introduced in 2023, has evolved into a scalable, open- source solution for Central Monitoring (CM), supporting reproducible, GxP-compliant risk signal manage- ment with over 50 releases across 10 packages and is now deployed across 20+ clinical studies. This paper introduces the{grail}R package (Gsm Risk And Issue Listing), which provides business logic for actioning risk signals and enables systematic and transparent oversight.{grail}demonstrates how analytics and business processes can be aligned to strengthen clinical trial quality monitoring | adverse events, data cleaning, clinical trial | source presentation source |
6 |
| 281 | {tflmetaR}: An R Package to Manage Metadata of Statistical Outputs | Gary Cao | clinical trial, csr, r package | source presentation source |
6 |
This site serves as a convenient hosting location and does not work as a publisher. All copyrights belong to their respective owners.
The information on this site is provided for general reference purposes only. Whilst every effort is made to ensure that the site is up to date and accurate, no warrant, nor any responsibility or liability for, the accuracy or completeness of the content or for any loss which may arise from reliance on information contained in this site is provided. Links to other sites are provided for your convenience but no responsibility or liability for the content of those sites or of any external site which links to this site is accepted. References to any specific commercial product, process, service, or corporation are also for information only, and do not constitute endorsement, recommendation or favoring.
SAS is a registered trademark of SAS Institute Inc. SAS, Statistical Analysis System, and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. ® indicates USA registration.