The future of clinical development: Opportunities for disruption
By: Sowmyanarayan Srinivasan, Mayank Anand and Abhay Jha
Background
With an increasing focus on improving efficiency and acceleration of clinical trials, it is important to review the overall clinical development process and identify areas of opportunity to disrupt current ways of working. Patient centricity continues to be a topic with minimal change on the ground — this is true across multiple areas of protocol design, access, participation and diversity.
Clinical trial execution and data management also present multiple opportunities for improving the efficiency, accuracy and overall success of clinical research. Industry is now realizing the importance of risk-based data management. While ideas like DCT and hybrid trials have been spoken about for a while, they have not seen enough adoption, nor execution with the right intent and the expected outcome. The pandemic kickstarted the disruption, but that has not kept pace.
With these thoughts lined up in the agenda, a set of senior clinical business leaders met to discuss the opportunities, what it will take to change the process and how technology can help facilitate the change. We have curated the takeaways from these discussions.
Value chain: Protocol to submission
The clinical value chain encompasses various stages, from protocol creation to study planning and execution until submission with patient care at its core. Pharma companies have started focusing on stronger collaboration, patient centricity, technology adoption and process optimization to automate and streamline various functions in the drug development process. Highlighted ones in the image below are areas where most of the business leaders feel there is opportunity for disruption leveraging digital and technology solutions.
In this paper, we deep dive into some of the above highlighted areas, from protocol design to database lock, where we see opportunities for better process transformation and technology adoption to increase patient experience and reduce study submission timelines.
Opportunities for impact
1. Designing the right protocol
The focus on protocol design has grown over the last few years with a realization that a “better designed protocol” can reduce a lot of challenges associated with trials, ranging from patient experience, site experience and data quality to faster submissions, to name a few in the value chain. Captured are some specific actions that can help in creating a more aware protocol:
a. The percentage of protocols that learn from previous protocols continues to be low. Less than 1% of historical trials data has contributed to the creation of new trial protocol; therefore we need to emphasize the value generated through learning from past protocols.
b. With an intent to capture more end points, protocol is becoming complex. Often 2–3 protocol designs get combined to create new protocol, leading to lower levels of site engagement, patient adherence and subsequent amendments.
c. Practice of merging 6–7 cohorts over the course of the trial, as laid out in the protocol design, complicates the data collection process, analysis and interpretation of results.
d. Lack of clear demarcation between inclusion and exclusion criteria in the protocol design leads to a reduction in patient enrollment and subsequent increase in protocol deviations.
e. On average, ~25% of the data collected in a trial to capture omics data is exploratory in nature. Industry currently lacks ways to process this data for creating comprehensive protocols with the right endpoints.
f. Practice of capturing digital biomarkers, assay data only, in the last visit is a challenge as it pushes the overall timeline by 10–12 weeks (this includes 4–5 weeks for sample to reach lab, and then the lab cultures the specimen for another 4–5 weeks).
g. Most protocols are not comprehensive in terms of elaborating on the use of eSources like EDC, Wearable, ePROs etc.
2. Ensuring patient centricity
Patient centricity has long been a key topic for not just clinical trials but also the pharma industry in general. Patient centricity is a multi-dimensional topic, cutting across increasing the patient pool that participates in trials, diversity, experience, burden, etc. Here we capture some of the key points that bubbled up in the session.
a. Access to right patient pool
i) Patient recruitment continues to follow a traditional path with only ~5% of the eligible patients participating in a clinical trial.
ii) Inclusion and exclusion criteria in clinical trials are at times too stringent to make room for patients to participate.
iii) Despite advancements in technology and digital solutions to improve recruitment outcomes, many sponsors primarily rely on traditional methods for patient recruitment.
b. Patient diversity
i) Underrepresentation of certain populations — e.g., racial and ethnic minorities, elderly individuals and individuals with disabilities — continues to be a critical issue in clinical trials that ultimately creates gaps in understanding how treatments may affect these populations.
ii) The representation of the patient population is generally 50–60% from the U.S., 25–30% from EU and the rest from APAC, and therefore there is a need to identify how diversity can be improved without complex protocol design.
iii) While digital technologies have the capability to reach a broader and more diverse patient population, there are challenges related to internet access, digital literacy and language barriers that prevent the widespread adoption of such tools.
c. Reducing patient burden
i) An industry challenge is that it is looking for the perfect patients who fit in a complex protocol. Reducing patient burden when dealing with complex biomarkers is an important goal in medical research. For therapeutic areas (TAs) like pediatrics, complexities like having additional biomarkers can lead to reduced patient participation.
ii) Factors like inflexible trial schedules, frequent site visits, travel and expenses reimbursement continue to be pain points for the patients and have not been addressed so far.
iii) eConsent-translated forms in local languages are very complex due to the regulatory mandate of “no alteration,” leaving patients with no other choice but to rely on lawyers to understand eConsent.
3. DCT as a methodology and an enabler as opposed to “just” technology
Decentralized clinical trials should be viewed as an operating model/way of working with the intent to make trials more patient-centric.
a) DCT should be viewed as a way of improving patient participation in clinical trials and driving patient centricity. Keeping in mind, for this approach, protocols should be reviewed in the context of making it more patient-centric, reducing patient burden while finalizing schedule of assessment.
b) Historically, most of the trials branded as “DCT” are dominantly pilot projects that combine conventional trial and decentralized trial approaches to assess if the results have a correlation or not.
c) The concept of DCT is usually introduced after the protocol creation, which leads to downstream inefficiencies in capturing trial data and an increase in protocol amendment.
d) DCT involves conducting clinical trial with decentralized approaches like remote monitoring of patients, electronic data capture, etc. However, technology supporting DCT is not planned well and integrated in the overall business process.
e) Sponsors hesitate to go to IRB and regulatory bodies because of associated compliance risks attributed to maintaining standards and scientific integrity.
f) The language of eConsent as provided to patient is getting difficult to comprehend and is counterproductive, resulting in a reduced number of patients who get enrolled.
g) With the latest compliance around patient privacy, diaries using emails of patients has created a challenge. Dummy ID for emails is an option but has seen less adoption so far.
h) Traditionally, clinical trials have been designed with a focus on the investigator’s needs. However, regulatory bodies, sponsors and technology firms must adopt ways of making trials more patient-centric — e.g., developing guidelines for the use of emerging technologies and ensuring that data collected remotely are reliable and secure.
4. Adoption of digital health in clinical setting
Pharma companies have recently been experimenting with digital technologies like eCOA, ePRO, EHR, telehealth and telemedicine platforms, etc., but the pace of adoption is slow, as the industry has historically been reluctant to embrace change. To fully explore the gamut of capabilities offered by digital tools and similar technologies, collaboration between stakeholders that include regulatory bodies, technology providers, sponsors and patients is critical. Below-stated pointers need to be investigated for a wider adoption of digital tech in clinical trials.
a) Steering through rigorous validation and documentation processes to include these digital technologies is complex and often time-consuming.
b) Safeguarding patient privacy for any potential data breaches demands foolproof solutions that are yet to make their mark in mainstream trials.
c) Lack of clarity on the ROI of implementing digital tools requires investment in technology, infrastructure and training.
d) It is often challenging to integrate digital tools with existing clinical trial processes and systems and requires detailed evaluation for the right fit.
5. Emerging data sources and role of EDC
With increasing number of data coming from outside EDC, the role of data sources and data managers is being redefined.
a) With ~75% of clinical trial data now collected outside EDC, data managers need a better way to manage clinical data coming in from diverse sources, at scale.
b) TAs like oncology demand image processing, for which external vendors (e.g., Mint) are looped in. The challenge is current EDCs do not offer capabilities to process images and only carry structured patient-level information.
c) Sponsors are lacking technology that can bring together EDC and image-processing capabilities through an AI/ML layer built on top for multiple use cases — e.g., a single repository for drilling down into trial-specific data and insights.
d) Innovation in TAs like neuroscience focused on Alzheimer’s and dementia has been a challenging task. Since these are complex multifactorial diseases, these trials rely a lot on third-party vendors who share the information in 15–20 different data types, resulting in a considerable amount of time spent in capturing and processing it.
6. Excellence in trial operations
Focus of the discussion was primarily on monitoring and an approach to bridge the gap.
Statistical data monitoring
a) Monitoring plans in clinical trials is usually brought very late in the process, which weakens the effectiveness of risk-based monitoring and site-monitoring activities. It has been observed that pharma companies have a central monitoring team and risk-based monitoring team operating in silos. What they currently lack is a statistical monitoring function that can act as a binding layer between these two functions by flagging the sites at risk and scheduling CRA visits as opposed to just looking at erroneous data and making decisions.
b) The trend of buying fancy technology solutions for RBQM and data management separately does not work well for the sponsors and calls for bringing together these two functions on a common platform.
c) Need for one overarching data scientist role looking into both site monitoring and data quality. To break these silos, data manager’s role needs to evolve to have capabilities of a data scientist.
7. Prioritized and risk-based clinical data review
Risk-based data review has been a key topic for a while and provides an opportunity to reduce submission timelines.
a) Pharma companies often spend a lot of time in data review of non-critical variables that increase the overall cycle times. Out of all queries raised, only ~2.5% causes primary or secondary data change; 95% of queries are raised on noncritical data points.
b) Critical variables need to be defined in the protocol itself for risk-based clinical data review. The choice of putting database lock on hold for non-critical variable queries depends on sponsors and a willingness to adopt risk-based approaches.
c) Protocols currently lack an inherent risk-based data review model that can inform both data management and risk-based monitoring teams of their data-review-specific responsibilities in the beginning of the trial.
8. Trial submission
The trial submission phase also has multiple opportunities to optimize timelines, but we couldn’t deep dive due to a lack of representation from a regulatory group, and we plan to cover in next iteration.
Recommendations
To address the above opportunities, there is a need to think across multiple dimensions — future business processes, ways of working, technology to enable change, regulatory impact and people change management.
Here we are highlighting the role of “technology to enable change,” especially the broad range of AI capabilities in the clinical value chain. Here are some approaches which, if executed, effectively can accelerate the clinical trial process.
Broad set of AI capabilities across automation, predictive modeling, NLP/NLG and the latest generative AI.
Use of generative AI: Gen AI has the potential to revolutionize the ways clinical trials are managed.
a) Document generation: Gen AI can assist in the creation of protocol documents, informed consent forms, monitoring and a statistical analysis plan by analyzing historical documents and regulatory guidelines. It can also support compiling regulatory submission documents.
b) Patient recruitment and enrollment: Gen AI can assist in refining and optimizing patient eligibility criteria, ensuring that they are both realistic and scientifically sound.
c) Decentralized trials: Gen AI can assist in remote monitoring of patients by analyzing real-time data from wearable devices and other monitoring tools.
d) Trial monitoring: Use of predictive analytics can help in monitoring patient safety by identifying patterns that may indicate adverse events or safety concerns.
e) Data management and analysis: A lot of use cases where standards need to be mapped along with identifying anomalies in clinical data can be easily solved by using generative AI. Gen AI can also help in statistical analyses of clinical data, including descriptive statistics, inferential tests and regression analyses.
