In June 2025, the U.S. Food and Drug Administration (FDA) launched the Electronic Language System Assistant (Elsa), a generative AI tool designed to assist FDA staff with clinical protocol reviews, adverse event summarization, and inspection target identification. For clinical data management teams, Elsa signals a shift toward AI-assisted regulatory scrutiny that rewards well-structured, consistently documented, and traceable data. This opinion paper examines what Elsa does and does not do, identifies implications for data management practices, and connects these developments to the broader governance landscape established by the joint FDA–EMA Guiding Principles of Good AI Practice in Drug Development (January 2026) and the ISPE GAMP 5 Second Edition framework. We propose practical steps clinical data teams can take in 2026 to strengthen inspection readiness and position their organizations for an increasingly AI-enabled regulatory environment.
SCDM’s mission has always been clear: to shape, empower, and equip the clinical data professionals of the future. Within this mission, Vision 2030 charts our path forward, and JSCDM plays a central role, serving as both scientific anchor and catalyst for the knowledge our global community depends on.
Introduction. In the realm of clinical trials, the Clinical Data Interchange Standard Consortium (CDISC) standards have become increasingly mandated by numerous national and regional regulatory agencies. For several years, the modeling of observational studies (OSs) was not explicitly addressed in the SDTM Implementation Guide (IG). However, in 2024, the document “Considerations for SDTM Implementation in Observational Studies and Real-World Data v1.0 (Final)” was released. Aim. This paper aims to describe the challenges encountered when applying the CDISC SDTM standard to map data from real-world studies before the release of the CDISC document dedicated to OSs. Methods. The SDTM mapping process began exploring the rationale for using SDTM datasets and continued through programming, and validation. Results. Data from three OSs conducted by IQVIA Solutions Italy between 2020 and 2024 and enrolling 1543 patients affected by asthma, diabetes and a rare disease were analyzed. A total of 86 SDTM domains were created, 95% of which were conventional SDTM domains according to the CDISC guidelines. Main validation issues arose because Exposure dataset was missing, or because variable EPOCH was not found or due to violation of conformance rules (e.g. “Value not found in codelist”), as these rules, designed for clinical trials had not been adapted yet to OSs. These issues were managed on a case-by-case basis through changes to the SDTM domains or by providing documented justification. Conclusions. The challenges experienced till 2024 can now be solved thanks to the document released for OSs. However, implementing SDTM for OSs still needs ad-hoc solutions.
Introduction The development of generative artificial intelligence (AI) has been driven by advances in AI and machine learning, leading to innovative applications across various fields such as natural language processing and image generation. Particularly in the clinical development industry, generative AI has contributed to streamlining data analysis and research support, thereby fostering progress in personalized medicine. However, the current situation is characterized by a lag in governance and regulatory compliance related to its use. Aim This study aims to investigate the current status of generative AI utilization and governance within organizations in Japan's clinical development industry, clarifying the extent of adoption and identifying associated challenges. Methods Between May 21 and June 13, 2025, an online survey was conducted targeting pharmaceutical companies, contract research organizations (CROs), academic research organizations (AROs), and system vendors. A total of 35 items were collected regarding organizational attributes, the status of generative AI use, governance measures, and training activities. Data were aggregated and analyzed using keyword analysis and word cloud visualization to identify salient features. Results Respondents and organizations involved included AROs (49 respondents across 36 organizations), pharmaceutical companies (48 respondents across 21 organizations), CROs (33 respondents across 13 organizations), and system vendors (2 respondents across 2 organizations). All respondents were individuals with responsible positions to make decisions regarding the use of generative AI within their respective organizations or departments in clinical development. Approximately 76% of respondents reported obtaining approval for generative AI use, with tools such as OpenAI's ChatGPT series and Microsoft Copilot being predominantly used. The status of governance varied between organizations; more than 83.3% of pharmaceutical companies (40 respondents), 60.1% of CROs (20 respondents), and 16.3% of AROs (8 respondents) had some form of governance documentation related to generative AI use. However, the development of operational-level standard operating procedures (SOPs) was insufficient across all organizations—only 8.3% of pharmaceutical companies, 6.1% of CROs, and none of the AROs and system vendors had such documents fully in place. Generative AI was mainly used for translation of documents, brainstorming, and document creation and maintenance, with expectations for future applications including advanced data analysis and programming tasks. Benefits cited included increased operational efficiency, automation, and creative support, while barriers such as security and privacy concerns and risks of misinformation were also noted. The focus of education and training centered on AI literacy and safe usage practices, emphasizing the need for strengthened security education within organizations. Conclusion We investigated and summarized the current status of generative AI utilization and governance for clinical development in Japan. A key finding was the lack of organizational governance documents related to utilization of generative AI and education and training. There is an urgent need to establish such governance frameworks along with more practical education and training methods.
A case report form is a document designed to record all of the protocol-required data on each participant in a clinical study. A case report form is one of the key clinical study documents and is a tool that allows the collection of accurate and complete study data. This chapter covers case report form design and development including data definition, mapping, and guiding standards.
Introduction:As clinical research transitions from traditional site-based models to decentralized clinical trials (DCTs), understanding participant experiences is essential. While DCTs promise improved accessibility and operational efficiency, empirical insights into how participants navigate these formats remain limited. Objectives: This study explored participant experiences in both traditional and decentralized trials to identify factors influencing engagement, satisfaction, and retention. A secondary aim was to generate recommendations for participant-centered trial design. Methods: We conducted online, semi-structured focus groups with individuals who had recently participated in either traditional or decentralized clinical trials. Transcripts were analyzed using inductive thematic analysis to identify key patterns of participant experiences. Results: Five overarching themes emerged: Navigating Trial Modalities, Drivers of Participation, Communication and Relational Dynamics, Structural and Psychological Gateways, and Technology in Practice. Decentralized trials were valued for flexibility and integration into daily routines, while traditional trials offered supportive in-person interactions, but posed logistical burdens. Across both models, participants consistently identified communication quality, trust in clinical relationships, and feeling respected as the most salient influences on satisfaction and retention. Conclusion: While trial modality shapes logistical aspects of the participant experience, relational and structural factors such as transparent communication, emotional readiness, and technology usability were more influential in determining engagement and retention.These findings underscore the importance of embedding participant-centered design principles in both traditional and decentralized trials. Enhancing relational engagement, clarifying expectations, and addressing digital and logistical barriers may not only improve participant satisfaction, but also optimize trial efficiency and data quality.
To meet the growing need for strong clinical evidence on a global scale, both publicand private sectors have invested to standardise health data elements and achievegreater connectivity and interoperability of health data systems. This is enabled byFAIRification (making Findable, Accessible, Interoperable, and Reusable) of clinicaltrial data increasingly recognised as a crucial step in enhancing the value and utility ofclinical data across the research community. This is even more relevant in thepaediatric and rare disease field where data remains highly fragmented in terms ofdata collection practices, ontologies, and clinical reporting standards, and often lockedin silos with diverse formats and standards.
Introduction: REDCapCloud.com is a data science platform designed for regulatory-grade usage which affords virtual consenting and electronic data capture utilizing branching logic to create customizable study workflow processes. REDCapCloud.com was selected to facilitate enrollment and data collection for a large-scale genomics study on resilience-related risk/protective factors in military personnel. Objective: Utilizing an Institutional Review Board (IRB)-approved protocol for our proof-of-concept scenario, we created a REDCapCloud.com-designed study to simulate and test the ability to: 1) Identify participants which meet the study's inclusion criteria; 2) Allow study participants to schedule a virtual consent call and sign consent and HIPPA documents; 3) Email consented study participants their signed electronic consent and HIPPA forms; 4) Deliver 15 web-based surveys and collect responses, and 5) Request their mailing address if study participants consented to provide a saliva sample for DNA analysis. Methods: We performed a hybrid moderated and unmoderated workflow study to assess the platform’s capability to pre-screen individuals, conduct virtual informed consent, and collect survey data. Results: REDCapCloud.com correctly screened mock study participants meeting inclusion criteria, successfully scheduled virtual consent, and effectively collected consent/HIPAA signatures and emailed participants password-protected forms. Programmed branching logic successfully triggered the sequential delivery of 15 instruments for data capture and collected shipping addresses for DNA analysis. Conclusion: REDCapCloud.com is a highly functional platform capable of expediting study screening, enrollment, virtual consent, and data capture in a graphical user interface for data entry with a validation component to check user data and a de-identification component to make data less identifiable.
Aim: A state-wide obstetric medicines information service (OMIS) is offered as a free telephone service by pharmacists at a tertiary women’s and newborn hospital. The methods for recording OMIS enquiries used included hard-copy tools, Microsoft Excel® spreadsheets, and Microsoft Access® databases, had significant limitations which hindered service delivery and user efficiency. The study aimed to develop an electronic tool which is easily accessible across various electronic devices, with secure data storage and access. Method: The database used by other medicine information services were explored. Electronic Data Capture (REDCap®), was deemed the most appropriate tool, as a secure, web-based application, was selected for its simplicity, and robust functionalities. Results: The REDCap® system was used to develop, test, and review the efficient capture and reporting on essential components of OMIS enquiries. 6 months post implementation, a user satisfaction survey was circulated to all pharmacists who deliver and supervise the service. The REDCap® feature considered most satisfactory were the simultaneous multi-user access functionality, ease of entering data and time efficient in data entry with drop-down selection. The integration of REDCap® and Microsoft Power BI® enabled continuous and efficient reporting of data collected. Data displayed on dashboard could be filtered to obtain relevant report effectively, including call count by caller type, calls received by month, patient type, and location of caller. Conclusion: The implementation of REDCap® and PowerBI® in data management in the medicine information service has significantly improved data accuracy, user satisfaction, and reporting efficiency.
As I look back at the first 5 months of the year, I am amazed at how much we have accomplished at SCDM! Before I get to our accomplishments, let me share our vision and priorities for the next five years and my focus as the 2025 SCDM Chair. Our 2030 Vision is centered on four key areas: “Clinical Research 2.0”, End-to-End Data Flow, Intelligent Technologies, and Patient’s Choice. In each of these areas, SCDM will identify opportunities in developing an education strategy to advance our profession from Certified Clinical Data Associate (CCDA), through Certified Clinical Data Manager (CCDM), to Certified Clinical Data Scientist (CCDS). Additionally, we want to grow and leverage our partnerships to bring to new ideas, technologies, and processes to SCDM and the industry. Through all of this we grow our global influence.
The integration of artificial intelligence and machine learning (AI/ML) in clinical trials offers immense potential to reshape drug development and research efficiency. This review explores the multifaceted landscape of AI/ML applications in clinical trials including seven use cases of AI/ML that aim at improving data quality and enhancing patient outcomes and clinical trial successes. We discuss good machine learning practices focusing on clear scope definition, transparent risk-proportionate management, and robust feature engineering. We also highlight privacy-preserving data-sharing techniques like federated learning and the role and the risk of using large language models in patient recruitment, data capture, clinical decision support, patient engagement, and trial design optimization. This review highlights the potential of aiding clinical trials through the responsible use of AI/ML, while recognizing the challenges (e.g., generalizability, transparency, and robustness) as well as ethical considerations including patient safety, privacy, and human rights. It also serves to guide interested parties towards the responsible and effective integration of these technologies with clinical trials.
Introduction: The HL7 FHIR standard is widely adopted for sharing healthcare information. FHIR definitional resources are being explored as a means to support research data collection from clinics to study sponsors. A study's schedule of activities (SoA) and activity details must be clearly defined for success. eCRF libraries of SoA activities are developed as sponsor or therapeutic standards for study data management. This work introduces a FHIR-based methodology using graph techniques to create and manage SoA activity libraries, enabling SoA specification in FHIR formats. Objective: This study aimed to (a) develop methodologies for creating and managing activity specifications using FHIR definitional resources and (b) build a library of standardized research SoA activities for implementation by EHRs or other applications. Methods: Python programming and graph methodologies were employed to develop utilities for creating, editing, and managing SoA activity specifications as FHIR resources. Definitions from publicly available sources like LOINC and SNOMED were transformed into graph representations for editing and library management and converted into FHIR resources for study SoA specifications compliant with the HL7 Vulcan SoA Project Implementation Guide. Results: A standardized graph methodology was developed to create and manage SoA activity requirements as FHIR resources. Libraries of common and sponsor-specific activity definitions were built using LOINC data domains to support study requirements. Conclusion: Graph methodologies for defining and managing SoA activity definitions were successfully implemented, generating FHIR interoperability resources. Libraries of standard data domains in FHIR format were created to support clinical research operations.
The transition from traditional to modern data collection, in the form of eSource, has spawned an entire new ecosystem of approaches and vendors. This has created a complex web of considerations that leave many clinical research stakeholders unsure of where to start. A study from the Society of Clinical Data Management (SCDM) eSource Implementation Consortium, published in Contemporary Clinical Trials Communications in December 2024, found there were 36 steps research sites needed to traverse before they could even get started with eSource- all of which were associated with their own challenges. 1 Scalable implementation relies on the clinical research ecosystem, in collaboration with regulators, coming together to overcome these barriers. To this end, SCDM hosted a series of workshops and meetings in 2024, bringing leading subject matter experts from across the clinical research ecosystem together to share their knowledge, insights, and case studies. The initiative included three working groups, which focused on site readiness, contract readiness, and technological readiness. They were followed by a roundtable discussion, which reviewed and discussed the working groups' findings, the challenges and, crucially, the potential solutions of eSource implementation. This document summarizes those talks and sets out a range of best practices, all aimed at building a consensus that can act as a foundation for the wide-scale adoption of eSource clinical trial data collection and transfer across the industry. 1 Cramer, A.E., King, L.S., et al (2024). Defining methods to improve eSource site start-up practices. Contemporary Clinical Trials Communications, 42, 101391.
ABSTRACT BackgroundThe United States (U.S.) Food and Drug Administration (FDA) conducts Good Clinical Practice (GCP) inspections to evaluate regulatory compliance, assess clinical trial conduct, and verify data integrity in support of marketing applications. To date, there has been no comprehensive assessment of the geographic distribution or extent of such inspections conducted during the marketing application review. MethodsWe conducted a retrospective analysis of FDA GCP inspection records involving clinical investigators (CIs), sponsors, and contract research organizations (CROs) associated with marketing applications submitted during fiscal years 2016 to 2018. ResultsA total of 1,275 GCP inspections were performed in support of 347 marketing applications during the study period. Inspections covered CIs (86%), sponsors (10%), and CROs (4%). On average, three CIs were inspected for each application. Most CI inspections (63%) were conducted within the U.S., with the remainder primarily in Europe. Of the 347 applications, 36% had sponsor inspections and 15% CRO inspections; the majority of these were conducted in the U.S. (>80%). Approximately 4% of U.S.-based and 2% of non-U.S.-based CIs were inspected. CI inspections covered about 10% of enrolled trial participants overall (14% U.S. vs. 8% non-U.S.) and included review of approximately 5% of source records (8% U.S. vs. 3% non-U.S.). ConclusionsWhile the total number of inspections and source records reviewed varied by year, the distribution of inspection types, participant coverage, and source document review percentages remained consistent. Most inspections were conducted in the U.S., where CI inspections also covered a higher proportion of participants and source data.
In Japanese academia, Clinical Data Interchange Standards Consortium (CDISC) standards have not progressed mainly due to resource issues. The Tohoku University Hospital Clinical Research Data Center has integrated these standards into operations since 2013. Our objective is to implement CDISC standards to enhance quality and efficiency, and to establish systems for Study Data Tabulation Model (SDTM) and Analysis Data Model (ADaM) creation for electronic data submission. This paper describes our CDISC activities, focusing on experience and lessons learned in investigator-initiated clinical trials and registry studies through 2023. We assigned dedicated personnel to consolidate CDISC knowledge and manage outsourcing, facilitating implementation through Data Management, Biostatistics, and Medical Information Management collaboration. Work time decreased significantly after our first in-house implementation. Challenges were addressed via our Quality Management System, resulting in standardized CRF templates. We continue exploring optimal CDISC utilization from multiple group perspectives to standardize, streamline, and automate processes across our organization.
The landscape of clinical trials is rapidly evolving, driven by advancements in Artificial Intelligence (AI) and the shift towards decentralized, hybrid models. This perspective explores two critical topics: 1. The role of AI in revolutionizing clinical trials while ensuring ethical standards.2. The utilization of real-world data (RWD) to enhance patient-centric research. Together, these themes highlight the potential for transforming clinical research into a more efficient, equitable, and effective endeavor.
As the clinical research landscape continues to evolve, so does our collective responsibility to shape its future with rigor, innovation, and inclusivity. JSCDM’s Fall 2025 Issue reflects the dynamic pulse of our field—where data standards, digital tools, and decentralized models converge to redefine how trials are designed, conducted, and scaled.
Objective: Research Electronic Data CAPture (REDCap) is a powerful web-based data management tool commonly used in academic research centers. Proper use, development, and execution are crucial in facilitating high-quality data capture, ensuring robust future analysis. While REDCap provides integrated tools for assessing data quality and status, these often fail to address the intricate demands of multisite longitudinal studies. Methods: We present a framework to optimize REDCap project development and introduce a Python-based data quality pipeline. Results: By focusing on strategic pre-production project design and implementing rapid-response quality assurance during post-production, we substantially improved the quality and accessibility of data for analysis. The Blackbox was first released in November 2024 for a double-blind clinical trial. A review of the three output files, Query Library Flags, Missingness Flags, and Missingness Summary revealed that there were 1949 queries, with the violations occurring between 85 and 500 days. Most of these were due to changes to the protocol or to missing branching logic (resulting in unnecessary queries). The study team addressed and resolved queries where able, and fields with missing branching logic were fixed. Black Box execution was applied against the corrected data set. A comparison between the two runs shows all queries were resolved, and there were no violations. Conclusion: This approach can empower other researchers to enhance both the accuracy and usability of data in complex research projects by leveraging REDCap's capabilities.
Introduction/Objectives We propose and conduct an infrastructure for a fully remote, decentralized Psycho-Oncology trial using smartphones. The data collection flow and mechanism are discussed. Methods The Decentralized Clinical Trial system was built by a multidisciplinary team of researchers, data center members, and vendors. We facilitated virtual trials through web-based systems and zero site visit, Web-recruiting, eConsent, ePRO, Apps, and Google Analytics. Results Virtual Clinical Trials enabled via technological levers spread and improved the efficiency of clinical trials. Considering the characteristics of the research, building part or all of the web-based system dramatically reduced the burden on patients and researchers. Conclusions We would like to emphasize the potential benefits of our novel strategy of conducting a fully decentralized clinical trial while it is always important to take every possible precaution to ensure that participants who are not proficient with digital technology are not disadvantaged.
Clinical trials are critical for advancing medical knowledge and developing new treatments, yet they often involve substantial administrative burdens that can impede progress and reduce efficiency. Effective clinical trial data management requires seamless participant engagement, accurate data collection, strict regulatory compliance, and coordinated efforts among multidisciplinary teams. These demands can strain resources and lead to workflow inefficiencies. This research paper explores the potential of large language models (LLMs), such as ChatGPT, to enhance clinical coordination by optimizing administrative processes in clinical trials. We assess the model's applications in streamlining standard operating procedures (SOPs), clinical data management, automating documentation, and supporting regulatory compliance. To ensure responsible implementation, we also examine key challenges related to ethical considerations, data biases, and safety concerns, while proposing strategies for mitigating these risks. Our findings indicate that integrating AI-driven solutions like ChatGPT can significantly improve operational efficiency, reduce administrative workloads, and allow clinical trial teams to dedicate more time to patient care and scientific inquiry. By leveraging AI responsibly, we can make clinical research more agile, adaptive, and focused on advancing medical innovation.
