Syndication in science: Curated collaboration

Abstract

Development and validation of digital measures require dedicated clinical studies, which can be conducted by a single study sponsor or a precompetitive collaboration. In this perspective, we propose an alternative model, data syndication, a curated collaboration, which foresees a technology provider being a founding member with biopharmaceutical sponsors and other stakeholders joining. Its main advantages are the speed of the study startup and the opportunity for real‐time data streaming.

Digital innovation in life sciences originally was met with high hopes — and a healthy dose of skepticism. ¹ An explosion of digital sensor innovation to collect health‐related data over the last two decades was driven by sensor miniaturization, high efficiency in data storage, and real‐time transmission, which fueled the hopes of solving many problems of clinical trials, such as limited data points collected inside the hospital walls not reflective of what happens in patients' day‐to‐day real‐life activities.

The rapid proliferation of smartphones, enabling quick, real‐time data synchronization through cloud environments, real‐time analytics, and user‐friendly visualizations accelerated the process. The first palpable result was the rapid adoption of wellness products, a.k.a. fitness trackers and smartphone apps, along with digital self‐awareness and lifestyle modifications (e.g., the Quantified Self). However, this explosion of recreational digital data usage was not mirrored in clinical research. Several reasons were identified to explain the lag: the regulated environment and outstanding regulatory questions, the need to match the long lifecycle of drug development, and especially important, the requirement for empirical evidence to substantiate drug labeling claims based on digital health technology (DHT)‐derived data. ² Evidence generation, on the surface, seems to be the easiest to address by conducting well‐designed and efficiently executed clinical studies for which the biopharmaceutical industry excels. However, it has proven to be one of the most difficult to achieve.

The ultimate goal of developing DHT‐derived measures is to accelerate the development of new medicines. The highest level of regulatory acceptance for any new outcome measure in the US is a qualification of a tool via the Food and Drug Administration (FDA) drug development tool qualification program. However, relatively few tools receive a qualification status as the supporting evidentiary package is extensive and often it takes multiple studies to build. A nimbler approach is to develop and validate fit‐for‐purpose measures that are used for internal decision making by sponsors, including proof‐of‐concept studies. ³ If evidence generated in such studies is used for regulatory purposes, the study design to validate DHT‐derived measures and its results can be discussed with regulators. For example, in the US Critical Path Innovation Meetings are designed to discuss proposed DHT‐derived measures and receive general advice that is product‐independent and with nonbinding recommendations. ² An alternative approach is to use an investigational new drug pathway to receive specific advice related to a particular drug development once the sponsor‐specific context of use is determined. ² In either case, validation is key to the use of such measures with confidence and, in the case of digitally derived measures, requires sensor verification, analytical, and clinical validation. ⁴ A significant distinction between DHT‐derived measures and more traditional drug development tools, such as biomarker assays or imaging technologies, is the need for human participant studies to establish technical performance. ³ , ⁴ Such studies can be conducted in healthy volunteers or subjects with disease, but they add complexity as compared to, for example, biomarker assays which can be analytically validated at the bench using archived biologic specimens or tissue imaging studies. ³ , ⁵

Two general approaches for creating evidence to support DHT‐derived measure use in drug development entail non‐interventional studies aimed at establishing technology performance that can be conducted by (1) a single sponsor and (2) precompetitive collaborations. Advantages of the first approach include full control over study timelines and the potential to keep the study design and results confidential until drug candidate approval. At the same time, the sponsor is fully encumbered by all the risks and costs in personnel and funding associated with embarking on a digital measure development journey, including digital measures selection, study design, study execution, and the analysis of the data. Moreover, the timing of DHT‐derived measure development and validation studies does not fit well into traditional drug development strategies as incorporating the validation experiments into drug interventional clinical trials comes with serious limitations. ¹ It also requires a partnership with a technology provider as most biopharmaceutical companies do not have in‐depth expertise in DHTs and the infrastructure for data collection. ⁶ The alternative approach is a precompetitive collaboration that aims to address shared challenges that cannot be met by a single entity. The goals can include developing standards/infrastructure, data generation/aggregation, knowledge creation, and product development. Specific requirements for DHT‐derived data collection are described in final FDA guidance on Digital Health Technologies for Remote Data Acquisition in Clinical Investigations. ⁷ Participants in precompetitive collaborations may include academic and industry scientists, government entities, foundations, patient advocacy groups, and the public. Collaborations involve open or limited participation in project execution and access to outputs. Eight models of precompetitive collaboration in biomedicine have been described, including open‐source initiatives, industry consortia for R&D process innovation, discovery‐enabling consortia, public‐private consortia for knowledge creation, prizes, innovation incubators/insourcing, and industry complementor relationships. ⁸ The advantages of precompetitive collaborations include risk and expertise sharing ⁹ which becomes increasingly important in the context of quickly proliferating digital technologies with old DHTs being phased out and new ones coming to the market. A precompetitive collaboration may pursue various goals involving different types of participants and structures ⁸ with a growing track record of success. ³ The main downside of precompetitive work can be the speed of launching collaborative studies. Depending on the collaboration structure and founding members' intentions, it may take time to achieve a certain level of membership with sufficient funding to design, execute, and analyze study results. Additional challenges include the need to reach a consensus among consortium members which may pursue somewhat different goals depending on the mechanism of action, maturity, and lifecycle needs of their respective internal drug candidate pipeline. Other considerations include the need for specialized infrastructure to combine and align high frequency data from multiple DHTs through direct access, cloud pathways, and expertise in monitoring patient adherence.

The alternative model to a single sponsor effort or a precompetitive consortium is a newly emerging syndication model, a curated collaboration which is a hybrid between a single sponsor study and precompetitive work, to undertake measure development and test technology performance (Figure 1). The main prerequisite of this syndication model is that a founding member has sufficient resources, both economical and scientific, to seamlessly stream data from patients to the study sponsors, and the experience to design and conduct the studies, thus off‐loading major risks of undertaking the effort from the broader membership community. In contrast to a single sponsor study, the syndication model foresees other interested parties joining the effort, providing scientific and technical input, contributing to study funding, and endorsing the results. The curated collaboration model addresses the main challenge of a long lead time to enroll enough members and reach a consensus on the study design. In essence, the syndication model combines the speed of a single sponsor study with the advantages of multi‐member expertise and sharing of risk. In the field of DHTs, the role of a founding member is initially best filled by technology service providers, unbiased by medicinal products under development, because these organizations have appropriate infrastructure to conduct digitally enabled data collection, have personnel with sufficient expertise and training in DHT evaluation/validation study conduct, and can provide a neutral collaboration space for multiple biopharmaceutical syndication partners. A comparison of all 3 models is provided in Figure 1.

FIGURE 1.

Key features of a syndication model as compared to single sponsor studies and precompetitive collaborations.

Historically, drug development and technology/device engineering were separated, non‐overlapping fields. ¹ However, DHT adoption for purposes of drug development induced cross‐fertilization between these areas. Biopharmaceutical companies have employed device engineers to facilitate the adoption of DHT‐derived measures; at the same time, drug developers are motivated to join the technology sector to understand the details of the data collection process and provide drug development expertise to make the DHT deployment process more efficient.

As with any collaboration model, the syndication approach has its own advantages and shortcomings. The main trade‐off of the speed of study design and initiation is that one of the syndication members takes both a leading role and a major risk. In the field of digital measures, this speed can be a critical parameter to generate evidence as DHTs evolve fast and many may be phased out and replaced with new models in the time span of a few years. The other syndication members can join the syndicate, potentially influence study design and execution, contingent on joining sufficiently early to have an impact; while all members can have access to the data and can contribute to the study analysis and interpretation of the results. While the founding member, often being a technology provider, takes a leading role, participation of drug developers in the syndicate is indispensable. Contributions of other syndicate members go beyond providing funding support: their input on study design, types of data to be collected, and data analysis are essential. With speed comes the necessity that alternative approaches be assessed and consolidated to ensure that the best potential measures and models are explored quickly. This can be offset by the opportunity through pooled resources to explore more approaches than would otherwise be practical with a single sponsor, but the natural diversity of multiple independent, single‐sponsor studies over long periods of time is foregone. The other requirement is that the data integration platform should be overseen by a neutral party, providing equal access to the data and study results. As DHTs become normalized in clinical studies, platforms integrating data from a variety of sources will become more common. Ideally, such a platform should be hardware and data processing algorithm agnostic, allowing flexibility necessary for keeping up with technology updates. A syndication leader, who is also a technology platform provider and possesses relevant technical expertise, brings the opportunity to leverage real‐time data streaming, giving members new ideas and developing new approaches on how to collaborate around the data as they arrive. This leadership role can expand to other Clinical Research Organizations as DHT capabilities are adopted more broadly. Not only does data streaming ensure that data quality is addressed continuously, but it also affords new opportunities for the community to come together, brainstorm, analyze, and interpret the data from the very first patient in to the last patient out. It is important to be aware that early looks at the data during study conduct may impact future observations so must be managed accordingly. This type of collaboration is reinforced through joint publications of analysis methodologies and findings, thus speeding the dissemination of evidence and measures. Publishing methodologies and results require transparency, such as providing a general description of data processing algorithms, which should be encouraged for the benefit of the scientific community.

The syndication model is being appraised in a Parkinson's disease (PD) progression study (NCT06219629). This is a multi‐center, longitudinal, observational study to determine the usability, validity, and reliability of digital assessments over a period of 12 months among participants with PD. The main study objective is to evaluate compliance and usability of components of the neuroscience toolkit which includes both point‐in‐time task‐based data collection and continuous monitoring by means of a DHT. Additional objectives include DHT‐derived measures content validity, criterion validity, construct validity, known group validity, and the ability of DHT‐derived measures to detect change alongside expected disease progression. The study was initiated by a technology company, which developed digital assessments of PD, with several biopharmaceutical companies joining the partnership to evaluate the performance of these technologies. The DHT‐derived measures were selected based on assessments included in the PD “gold standard” Movement Disorder Society Unified Parkinson Disease Rating Scale that are amenable to remote digital data collection and are safe for patients to perform at home. The study protocol was developed by the technology provider with input on study design, data to be collected, and the analysis plan from biopharmaceutical partners.

While the curated collaboration model provides the advantage of a quick and efficient study startup for DHT‐derived measures, it is by no means limited to digital innovation. It is conceivable that this model can be applied to any collaborative effort that would benefit from the contribution of multiple stakeholders, emerging data sharing, analysis, and interpretation, requiring, at the same time, a neutral platform for data collection and distribution. Other examples include rare disease natural history studies, particularly if multimodal data collection is required, and AI‐model training on continuously updating data sets. In a recent discussion paper, FDA indicates that AI/ML has the potential to inform the design and conduct of clinical trials and provides examples of leveraging DHT‐derived data to predict the status of a disease and a response to treatments as well as analyzing large and diverse data set generated by DHTs, including multimodal data and composite measures. ¹⁰ Such data sets can also serve as synthetic study control arms, allowing sponsors to make comparisons with emerging treatments in their respective pipelines, often critical for rare disease interventional studies. Time will show whether syndication can be replicated across the industry and whether it presents a viable alternative to both biopharmaceutical sponsored and precompetitive collaboration models.

FUNDING INFORMATION

No funding was received for this work.

CONFLICT OF INTEREST STATEMENT

DH, ESI, and JAW are employees of Koneksa Health and may own company stock. All other authors declared no competing interests for this work.

DISCLAIMER

As Editor‐in‐Chief of Clinical and Translational Science, John A. Wagner was not involved in the review or decision making process for this paper.

Hurry DB, Izmailova ES, Davies S, Harari O, Stoch SA, Wagner JA. Syndication in science: Curated collaboration. Clin Transl Sci. 2024;17:e70085. doi: 10.1111/cts.70085