Insights from PHUSE US Connect 2024
ICON recently participated in the PHUSE US Connect Conference, held February 25-28 in Bethesda, Maryland.
PHUSE, the organiser, is a global non-profit organisation serving the life sciences industry. Run by volunteers with industry expertise, PHUSE fosters collaboration across data management, biostatistics, statistical programming, data science, and eClinical IT. Importantly, PHUSE also advocates for the industry, representing its voice to regulatory agencies like the FDA, EMA, and CDISC. The conference brought together 677 attendees from 173 companies across 18 countries.
The conference had an opening keynote followed by 6 concurrent sessions grouped into 17 distinct topics and 9 primary conference themes:
| Topics | Themes |
| Analytics and Statistics | AI for Clinical Data Scientists |
| Analytical Risk Based Monitoring | Digital Data Flow |
| Data Handling and Engineering | End-to-End with R Packages from SDTM to TFL |
| Data Standards and Governance | Global Leadership and Team Excellence: Navigating Borders, Minds and Success |
| Data Visualisation | Insights into Risk-Based Monitoring Practices |
| Emerging Trends | Navigating Open Source in the Clinical Data Space |
| Leadership | Oncology Programming Excellence: Strategies from RECEIST to ISE |
| Open-Source Technologies | Real-World Insights: Mapping, Scaling and Integration |
| People Leadership and Management | Tools for Data Exploration and Reporting Advancements |
| Poster Presentations | |
| Professional Development | |
| Real-World Evidence | |
| Scripts, Macros and Automation | |
| Software Demonstrations | |
| Standards Implementation | |
| Submissions and Agencies | |
| Therapeutic Area Specific Topics |
In addition to the above topics, there was a dedicated stream for the PHUSE/FDA Innovation Challenge. Participants were invited to work on prototypes and to present their proof-of-concepts or solutions.
This year, the theme was "Application of AI/ML Methodologies", with presentations falling within the following two pillars:
- Risk Based Monitoring
Throughout all stages of the clinical investigation, managing both risks to participants (e.g. safety problems) and to data integrity (e.g. incomplete and/or inaccurate data).
- Clinical Trial Developments
Development across clinical trial activities, such as data collection, data comparison and/or data analysis practices.
The FDA, recognising the potential of AI/ML to revolutionise new drug development, sought collaboration with the industry to gain insights into the latest applications of this rapidly advancing technology. A competition was held, with winning ideas chosen for further development by PHUSE working groups in partnership with the FDA.
ICON presented a poster and a paper on the following topics of special interest:
Poster presentation by Elena Valkanova and Bulat Fatykhov: "Comparison of Standard Log-Rank Test with Selected Alternatives"
In oncology studies with time-to-event outcomes, the hazard and survival functions are two ways of specifying a survival distribution. Using the constant hazard assumption for the exponential distribution could provide misleading conclusions. The clinical trial designs with two groups based on proportional hazards (PH) assumption are likely to be underpowered to detect treatment effects. When the PH assumption is violated, the likelihood estimator depends on underlying censoring time distribution.
Selected alternative methods for survival analysis when the assumption for the proportional hazards is questionable were compared and assessed the impact of sample size. Three selected methods – log-rank (LR), max-combo (MC) and restricted mean survival time (RMST) – applied to both simulation data and real clinical trial data, were presented with graphical presentations using R.
Shabeeb Kannattikuni presented a paper co-authored by Karthika Bhavadas: "A SAS Tool for Generating Adverse Event Summaries in XML Format Using ADaM Datasets, ADSL and ADAE for ClinicalTrial.Gov and EudraCT Submissions"
Clinical trial submissions to regulatory authorities require accurate and standardised adverse event summaries in XML format. Shabeeb presented a novel SAS® tool designed to streamline the process of generating comprehensive AE summaries using the ADaM while producing output in XML format adhering to the required schema and data standards for submission to ClinicalTrials.gov and the EudraCT.
To assess the performance and accuracy of the SAS® tool, a validation study was conducted using multiple real-world clinical trial data. The results demonstrate that the tool significantly expedites the AE summary generation process, reducing manual effort and minimising the risk of error in XML creation. The generated XML files conform to the industry-standard data structures and are readily suitable for direct submission to ClinicalTrials.gov and EudraCT platforms, promoting data integrity and expediting regulatory review.
Conclusion
The PHUSE conference offered a wealth of opportunities for attendees to connect and share knowledge on a range of technical subjects. A significant emphasis was placed on practical applications of Generative AI and various programming techniques using popular open-source languages such as R and Python.
The diverse audience, which included representatives from CROs, pharmaceutical companies, biotech firms, and technology vendors, fostered a valuable exchange of experiences and current implementations. This environment facilitated the creation of new relationships that hold promise for future collaborations in both business and technology advancements.
Kent Letourneau