Multiscale dosimetric and radiobiological modeling for radiopharmaceutical therapy (RPT), Part 4: digital twins for optimized RPT and their ethical and regulatory dimensions

Introduction: Radiopharmaceutical therapy (RPT) uses pharmaceuticals or devices generally labeled with particle-emitting radionuclides to treat metastatic cancer. In common practice, the administered activity follows the "one size fits all" approach, regardless of the patient-specific biodistribution of radioactivity. As highlighted in the previous three exhibits, to move beyond this paradigm, multiscale computational phantom modeling of dosimetry and radiobiology is needed. The long-term goal is to parameterize these phantoms with target and patient-specific characteristics to create digital twins (DTs): digital representations of individual patients that are operable (i.e., enabling virtual treatments to personalize therapies and activity prescriptions). Collected high-quality images, adequately annotated, combined with additional patient-specific information, and the multiscale bio-dosimetry, might lead to rapid deployment of DTs via innovations in artificial intelligence (AI). However, DTs also introduce a number of legal risks and ethical dilemmas that must be resolved.

Methods: A conceptual outline of ethical, socioeconomic, and regulatory dimensions relevant to applying DTs to RPT is presented.

Results: 1) DTs harvest a large quantity of real-time data, including personal information that should be kept confidential. However, although current legislation restricts non-essential data sharing, DTs will benefit from including as much data about the patient as possible. Seeking ethical approval at an early stage to allow patient informed consent for data sharing of prospective data should be strongly encouraged. 2) Worldwide differences in privacy regulatory frameworks and levels of restriction represent a major impediment. A possible strategy for dealing with this obstacle is the adoption of federated learning approaches to enable multi-institutional training of models, secondary data extraction, or data augmentation following detailed guidelines. 3) Guidelines on how to construct DTs for RPT based on the temporal activity distributions-e.g.,utilizing physiologically based pharmacokinetic (PBPK) models - and how to store and share the models (and their inter-/intra-organ relationships) still require to be developed. Scientific associations developing such guidelines should identify possible DT risks and their mitigation. 4) DTs used to make decisions need to be updated based on evolving data sets. Data feeding into the DT must be accurate, complete, and updated to ensure reliable output. 5) A significant economic barrier to long-term clinical implementation and development of DTs in RPTs is the current lack of reimbursement for sequential quantitative imaging and data processing to perform treatment planning and post-treatment verification of the absorbed doses. 6) Two critical open questions that need to be addressed include: "Who is liable for the decisions made using DTs?" and "How can we guarantee the interoperability of DTs, which are connected and interact with other DTs, creating a complex and multilayered system?" 6) DTs store vast quantities of valuable data, making them attractive to a data hack or a cybersecurity breach. Appropriate safeguards and security systems must be implemented and regularly updated to evolve with the technology. 7) DTs may source data from different geographic zones and interact in different jurisdictions. A common global framework that allows cross-border data sharing to reduce complexity in an already complex risk profile must be developed. 8) Finally, any additional potential risks should be identified and mitigated early on, to establish contracts with stakeholders flexible and adaptable to the changing technology, thus reducing barriers to DT implementation.

Conclusions: We present the concept of DTs as a promising tool to improve RPTs. Here, we highlight the potential barriers for DT development and implementation and motivate the community to work on solutions to fully leverage DTs for optimized RPTs.