Enabling Democratization of Clinical Trial Data Access Within Novartis

The contrast between the clinical trial data’s insight value and its high sensitivity means that a clear policy is needed to manage the opportunity and risk of re-using clinical trial data for secondary analysis. Furthermore, a data science platform that intended to serve the needs of hundreds of Novartis data scientists needed to find a scalable solution that wouldn’t bog down the user or the organization in inefficient bureaucracy. This article explores how data42 explored this problem and found a solution to it through the creation of a scalable governance process.

Many biopharma companies have an interest in enabling their R&D organizations to enable broad use of Clinical Trial data in a compliant manner. At Novartis’ data42 program we have made this one of the primary focus areas of our program’s data curation and unification efforts. You can read more about our endeavors in this pursuit, spanning from Achim Plueckebaum (leadership) to Anisah Alyahya (preclinical data) and many others.

It is no surprise that these types of technology investments have a clear value to R&D. Clinical trials are, after all, one of the most insight-rich data assets inside of any biopharma’s R&D unit. This is due to the fact that clinical trials are:

- performed on humans, as opposed to animal or cell-based studies

- focused on outcomes (as opposed to Real World Data Sources such as Claims or Medical Records datasets which rarely have such clear annotation of how patients respond to treatment.;

- for the most part, controlled — using random assignment of patients to experimental treatment vs. a comparator treatment (or placebo). Therefore, researchers across the industry are excited to identify opportunities to enable secondary analyses of such clinical trial data, as it allows for ‘secondary analysis’ for generating insights beyond the ‘primary analysis’ performed when the clinical trial was originally conceived of and executed.

Those points acknowledged, it’s also clear that clinical trial data are quite sensitive. It is widely appreciated that each datapoint should be considered precious since they are derived from patients who granted our clinical trialists and their physicians time and attention to participate in clinical research. Beyond time and attention, patients have consented to participate in our studies and for us to collect & processing their data. Furthermore, any improper use of data (e.g., performing ill-defined analytical approaches) may delay or derail clinical development or regulatory review — and thus actually impede the very same R&D processes we seek to enable).

Given the need to simultaneously safeguard patients’ rights and enable broad internal usage, data42 has placed a high value on a defined, universal, and robust internal policy for how we manage secondary use of clinical trial data.

Historically, Novartis R&D associates would adjudicate on a case-by-case basis individual data requests, which was often a cumbersome and resource-intensive process. We knew that these processes would not scale to data42’s vision of enabling hundreds of data access requests per year. An essential design feature of our work was the need to ensure scalability of the process so that requests for sensitive data access could be adjudicated with limited organizational overhead and resource consumption.

To build this policy we required the input of many functions and groups across the Novartis enterprise. We built a data access authoring committee that represented numerous functions including data privacy, compliance & ethics, medical affairs, legal, quality, regulatory, IT security, clinical development, and research representatives. After a year of discussions, meetings and considerations, we arrived at an agreement for a solution with multiple highly promising features.

The policy is based on evaluating the sensitivity of the clinical trial data being sought by considering (A) the phase of the trial data, (B) the stage of the program’s development, and © the category of question(s) being asked of the data. Highly sensitive access requests would be routed through technology-enabled human evaluation processes, whereas low or medium-sensitivity requests could be automatically adjudicated based on a clearly defined research plan.

Fortunately, we have since had the chance to even implement the policy and see how it performed in practice. After completing the first 500 data requests in a bit less than its first operational year, there are several noteworthy results and benefits of the policy. These include:

- Nearly fully automated through metadata interactions and requires limited human decision-making (<5% of requests were deemed high-sensitivity).

- Rapid access to data within minutes.

- Transparency to the data requestor — to help them anticipate the sensitivity of their requested data.

- Auditable and tracible to confirm that data were not improperly used.

- Support of good data science practice by requiring users to define their intended use of data when requesting access.

Having presented our work to other biopharma companies across the industry, we’re aware that having a policy gap between the technology’s potential value and the organization’s comfort has hindered many companies seeking to enable broad secondary use of clinical trial data. You can read about the details of the process of authoring and shaping this work in Drug Discovery today. In sharing this article I want to acknowledge the work of not only the co-authors but also the broad set of stakeholders across Novartis and our executive team members that made it possible. It is our hope that our policy work can provide an initial blueprint for the entire industry to enable the use of clinical trial data more broadly.

Ultimately nearly all of us will find ourselves in a position of compromised health at some stage of life. At data42, we believe it is a fundamental societal obligation to make rapid headway in discovering and developing new safe and effective treatments. We hope that this policy and the data-science enabled innovations that will emerge from it, may play a small role in accelerating life science R&D for all.

Lastly, thank you to the supporting members of the Novartis community who made this work possible, including the co-authors of the article referenced above, other members of our drafting and review committees and the Novartis Executive Committee members who ultimately endorsed this work.