Decentralized Clinical Trials in Oncology–A Radiological Perspective

WHAT IS A DECENTRALIZED CLINICAL TRIAL?

Conventional oncology trials typically require patients to travel to research centers for protocol-specified procedures. In contrast, decentralized clinical trials (DCTs) redistribute many of these procedures to qualified local facilities closer to patients. DCT is a design/model in which procedures take place outside conventional research sites and are supported by digital platforms, remote monitoring, and patient-centered logistics. By reducing geographic barriers and enabling remote participation through telemedicine and digital tools, DCT models have the potential to shorten trial timelines, improve retention, and shift research toward patient-centered frameworks. The COVID-19 pandemic accelerated the adoption of selected decentralized elements, although subsequent analyses suggest that DCT implementation has remained uneven across therapeutic areas and trial procedures [1]. In oncology trials, this transition is particularly relevant to imaging, where protocol-specified scans may be redistributed across a broader network of qualified imaging sites (Fig. 1). This expansion parallels the development of ambulatory oncology centers that co-locate imaging, laboratory, and infusion services at a single local care node [2].

Fig. 1. Expansion of the imaging source landscape under a DCT design. A: In the traditional model, oncology trial imaging is concentrated at a limited number of designated academic or tertiary imaging sites, and patients often travel to these centers for protocol-specified imaging. B: In the DCT model, qualified community imaging centers can be incorporated as imaging source sites, while image transfer, QC, and response assessment are coordinated through a cloud-based imaging platform and a virtual imaging core. This expanded network may improve trial accessibility and reduce patient travel burden but also would increase the need for prequalification of imaging sites, acquisition coordination, DICOM metadata monitoring, and remote QA. DCT = decentralized clinical trial, QC = quality control, DICOM = Digital Imaging and Communications in Medicine, QA = quality assurance, BICR = blinded independent central review, PACS = picture archiving and communication system.

WHY IS IMAGING DIFFICULT TO DECENTRALIZE IN ONCOLOGY TRIALS?

Imaging decentralization is difficult because oncologic response assessment requires not only image transfer but also reproducible acquisition, reconstruction, measurement, and longitudinal comparison across time points. Conceptually, the most consequential shift introduced by DCT in oncology is in the imaging sourcing model. Conventional oncology trials often centralize imaging activities at a limited number of academic medical centers equipped with dedicated research infrastructure, requiring patients to travel to these centers for protocol-specified imaging. Under the DCT design, qualified community imaging centers serve as additional imaging sites, while academic medical centers function as central hubs that house reading expertise. A cloud-based imaging platform, often implemented as a trial-specific cloud picture archiving and communication system (PACS), provides a unified environment, and a virtual imaging core supplies the centralized quality oversight that previously resided physically within academic centers.

Despite these enabling technologies, the adoption of decentralized elements has been uneven across oncology trial components. In registrational oncology trials supporting U.S. Food and Drug Administration (FDA) approvals during the COVID-19 period, remote site monitoring, telemedicine, remote laboratory assessment, and remote investigational product distribution were commonly adopted [3].

KEY COMPONENTS OF DCT IMAGING

Several radiology-specific concepts surface in DCT discussions, each describing a distinct operational layer (Table 1). These include imaging-site prequalification, cloud-based imaging platform, blinded independent central review (BICR), virtual imaging core, and imaging harmonization [4]. Site prequalification ensures that each participating imaging site meets a set of predefined technical requirements, including imaging equipment, acquisition protocols, and image transfer, before enrolling patients. BICR provides an independent and blinded assessment of imaging endpoints performed by central readers to ensure objective and consistent response evaluation. Unlike in a conventional central imaging core laboratory setup (where imaging is performed at a single physical location), in a virtual imaging setup, site qualification, quality control, and coordination of central image review are conducted through a cloud-based, geographically distributed infrastructure. Together, these elements form the operational backbone that allows imaging endpoints to be reliably collected and interpreted across geographically distributed sites.

Table 1. Key terminology in DCT imaging.

Term	Definition
DCT	A trial in which study procedures take place outside traditional research sites, supported by digital platforms, remote monitoring, and patient-centered logistics
Site prequalification	Pre-activation acceptance testing of imaging equipment, acquisition protocols, and data-transfer pathways at each participating site
Cloud-based imaging platform (Cloud PACS)	A web-accessible image archive that allows distributed sites to upload, view, and review images within a unified environment
BICR	Off-site, masked endpoint adjudication performed by readers independent of the local site and treatment assignment
Virtual imaging core	A remote operational hub that coordinates site qualification, quality control, and central reads in lieu of an on-site core laboratory
Imaging harmonization	Process of aligning acquisition parameters, reconstruction algorithms, and reporting conventions across heterogeneous sites
DICOM metadata monitoring	Automated tracking of header fields such as scanner model, reconstruction kernel, slice thickness, and contrast phase to detect protocol deviations
Adjudication rate	The proportion of cases requiring a third reader because of discordance between primary central readers

DCT = decentralized clinical trial, PACS = picture archiving and communication system, BICR = blinded independent central review, DICOM = Digital Imaging and Communications in Medicine

TECHNICAL AND REGULATORY CHALLENGES

Imaging variability is the central technical challenge in oncology DCTs. Multivendor CT systems use proprietary reconstruction algorithms and calibration practices that produce measurable differences in Hounsfield unit values, spatial resolution, and voxel texture, even under standardized protocols. Systematic discrepancies of 5%–15% in volumetric measurements have been reported across reconstruction algorithms [5,6,7], which can influence the Response Evaluation Criteria in Solid Tumors (RECIST) classification in borderline cases. Radiomic features and artificial intelligence (AI)-based response prediction models are even more sensitive to these acquisition-level differences [8], and phantom-based calibration cannot fully harmonize reconstruction-dependent variability.

Centralized reading addresses interpretive variability but does not eliminate it. Local–central discordance rates of 23%–46% have been reported in RECIST-based phase 2 trials [9], reflecting inconsistent lesion selection, variable measurement tools, and differences in reconstruction kernels or slice intervals. Central review thus standardizes interpretation while leaving acquisition harmonization as an ongoing operational task.

Cloud-based imaging platforms provide a shared environment in which participating sites upload trial images, imaging cores perform quality control, and central readers assess response using standardized criteria (Fig. 1). Integrated automated tools track Digital Imaging and Communications in Medicine (DICOM) metadata and flag protocol deviations before BICR. Recent evidence suggests that AI can further enhance image quality assessment in this workflow [10], although regulatory frameworks are yet to approve AI-reconstructed data or AI-assisted metrics for trial use.

Regulatory guidance has advanced but remains limited in terms of technical specificity for imaging [11]. The FDA’s Digital Health Technologies guidances (2021, 2023) and the Conducting Clinical Trials with Decentralized Elements guidance (2024) establish general principles of patient safety, data integrity, and protocol compliance; however, these guidelines do not specify the measurable imaging criteria. Harmonization profiles from the Quantitative Imaging Biomarkers Alliance (QIBA) and the American College of Radiology (ACR) remain voluntary and unendorsed by regulators [12]. Therefore, sponsors define their own imaging quality control parameters, accommodating technical diversity but producing variation across trials.

AN EXAMPLE: THE ALPHA-T TRIAL

The ALPHA-T trial (NCT04644315) is a phase 2 home-based decentralized trial of alectinib in patients with anaplastic lymphoma kinase (ALK)-positive solid tumors [13]. In this trail, while home visits were supported by mobile healthcare professionals and telemedicine, tumor imaging was performed at local radiology or imaging facilities. The protocol required patients to use the same facility throughout the study, staff to be instructed in RECIST v1.1 or Response Assessment in Neuro-Oncology (RANO) methodology, and CT and MRI parameters to meet predefined requirements, with images submitted to BICR [13]. As with other DCTs, conventional imaging infrastructure was not bypassed in this trail; images were captured at the qualified local facilities while preserving centralized oversight, protocol-defined harmonization, and independent review [14].

WHAT RADIOLOGISTS SHOULD KNOW?

For clinical radiologists at participating sites, DCT does not mean reading from home; DCT model implies operating within an institutional framework that exchanges images and metadata with a centralized infrastructure. Practical implications include the following: 1) adherence to protocol-specified acquisition and reconstruction parameters becomes more consequential than in routine practice, because deviations are automatically flagged, 2) DICOM metadata completeness directly affects eligibility for endpoint analysis, 3) cloud platform submission workflows and BICR adjudication pathways are increasingly part of the radiologist’s responsibilities, and 4) radiologists practicing at community imaging centers, who previously had limited exposure to clinical trial imaging, may increasingly find their practices involved in trial workflows. Familiarity with these operational concepts will become a baseline competency for radiologists involved in oncology trials.