We conducted a feasibility evaluation of a structured training program embedding trainees within a supervised human-in-the-loop workflow for oncology clinical trial data curation in support of the LookUpTrials digital platform (Figure 1). The primary objective was to assess whether individuals without prior clinical trials or informatics experience could be onboarded into standardized data stewardship workflows with appropriate supervision and quality assurance. The evaluation focused on workflow feasibility and participant experience rather than clinical adoption or patient-level outcomes.

A total of 10 undergraduate trainees from a university-based student organization were recruited through a competitive application process based on interest in clinical research, availability for longitudinal participation, and commitment to a defined weekly time allocation. Participants had heterogeneous academic backgrounds and varying levels of prior research experience, reflecting the study’s intent to assess the trainability of standardized workflows among individuals without formal clinical or informatics training.

Trainees underwent structured onboarding that included orientation to oncology clinical trial concepts, navigation of public trial registries and institutional trial portals, and use of standardized data abstraction templates. Training materials included written documentation, example entries, and step-by-step workflow guides. Trainees were embedded within a supervised, multitiered human-in-the-loop workflow that incorporated peer review and faculty oversight to ensure data accuracy, consistency, and completeness. Iterative updates to training materials and workflow documentation were made based on trainee feedback and observed sources of error.

Trainees curated publicly available oncology clinical trial information from institutional clinical trial portals and ClinicalTrials.gov across 5 academic medical centers. The sampling frame included all oncology clinical trials listed as active or recruiting on participating institutions’ public trial portals during the study period. Trials were selected systematically based on portal availability and recruitment status rather than through random sampling. Data collection was limited to 5 academic medical centers selected based on the availability of public institutional trial listings and existing program partnerships. The number of trials curated per institution was not intentionally balanced and varied according to institutional trial volume and portal structure.

Institutional clinical trial portals were used to capture site-specific information, such as local recruiting status and institution-level trial availability, which is not consistently available or timely in ClinicalTrials.gov. These data are necessary to support the institution-specific trial discovery functionality of the LookUpTrials platform. Extracted data elements included trial identifiers, disease site, eligibility descriptors, phase, recruiting status, and site-specific information. Data were entered into standardized templates designed to support downstream integration into the LookUpTrials platform. Quality assurance included peer review of entries and secondary verification by program leads to resolve discrepancies and clarify ambiguous protocol information.

The primary feasibility outcomes included trial throughput (number of trials curated over time) and descriptive changes in individual processing rates across the study period. The secondary outcomes included participant-reported experiences with onboarding, workflow clarity, peer support, and perceived learning, assessed via postprogram surveys with Likert-scale and open-ended items. Given the exploratory nature of the study and small sample size, analyses were descriptive and focused on characterizing workflow performance and participant engagement rather than formal hypothesis testing.

This study was conducted as an educational program evaluation and did not meet the definition of human subjects research. The project evaluated a training workflow and analyzed program-level feasibility outcomes using aggregated data and deidentified participant feedback. No identifiable private information was collected, and no patient-level data or protected health information were accessed. All clinical trial information was obtained from publicly available registries and institutional trial portals.

In accordance with US Department of Health and Human Services regulations for the protection of human subjects (45 Code of Federal Regulations 46), this work did not require institutional review board review or approval [22]. As the study did not involve human subjects research as defined by these regulations, informed consent was not required. Participant responses were collected voluntarily and analyzed in aggregate, and no financial compensation was provided.

Participant demographic characteristics are summarized in Table 1. In total, 10 undergraduate trainees participated in the program across the study period. Participants represented diverse academic majors within the life sciences and had heterogeneous prior research experience. Several participants reported no prior exposure to oncology or clinical trials at baseline. This heterogeneity aligned with the study objective of evaluating the feasibility of onboarding individuals without formal clinical or informatics training into standardized data curation workflows.

The number of curated oncology trials by institution is summarized in Table 2. Over a 10-month period, trainees curated a total of 2503 oncology clinical trial entries sourced from publicly available institutional trial portals and ClinicalTrials.gov across 5 academic medical centers. The volume of curated trials varied by institution, reflecting differences in the size and scope of publicly listed oncology trial portfolios. Curated trials spanned multiple disease sites and trial phases, providing broad coverage of oncology clinical research activity represented within the source portals (Table 3).

A descriptive analysis of individual trial processing rates indicated increases in throughput over time for most trainees (Figure 2). These changes were consistent with increasing familiarity with trial structures, data abstraction conventions, and workflow expectations rather than formal performance benchmarking. Variability in processing rates across trainees was observed, reflecting differences in baseline experience, availability, and learning trajectories.

Throughout the study period, peer review and secondary verification by program leads identified common sources of ambiguity in trial protocols, including inconsistent terminology across institutional portals and variability in how eligibility criteria and recruitment status were reported. Iterative refinements to training materials and workflow documentation were made in response to these challenges. Over time, fewer clarifications were required per curated trial, suggesting increasing alignment with standardized abstraction conventions and quality assurance processes.

Postprogram surveys indicated that participants perceived structured onboarding, shared documentation, and peer support as helpful components of the workflow. Participant ratings of select program components are shown in Figure 3. Trainees reported increased familiarity with clinical trial structure, eligibility terminology, and the role of structured trial information in supporting digital clinical research tools. Participants also noted challenges related to interpreting heterogeneous trial descriptions across institutions and managing time commitments alongside academic responsibilities (Textbox 1). Qualitative feedback suggested that the supervised, iterative workflow design supported learning and engagement despite these challenges.