Top pharmaceutical companies sponsor over a hundred clinical trials annually, generating vast amounts of data. Harnessing this deluge is a monumental task. eClinical Solutions, led by CEO Raj Indupuri, tackles this through advanced applications of data analytics and machine learning with a strong emphasis on AI in clinical trials optimization.
Specifically, eClinical Solutions taps AI/ML for automated data mapping, classification, review and mining insights. This enhances efficiency, speeds up cycle times and ensures quality as data complexity grows. The company’s Elluminate platform integrates and structures data, supporting advanced analytics.
Powerful techniques like anomaly detection algorithms can automatically flag potential data issues for human review. ML also classifies and categorizes data to focus reviewer time on the most critical areas. Fostering an innovative culture open to new ideas is key, too, he said.
In the following interview, Indupuri explains digitization has expanded data variety and volume. AI and ML are thus now critical to effectively harness this data and optimize trials. But thoughtfully implementing these technologies requires overcoming challenges like change management and ensuring model interpretability.
Can you share how eClinical Solutions is using data analytics and machine learning in clinical trials?
Indupuri: At eClinical Solutions, we provide an enterprise life sciences software platform and biometrics services. Our clients and biometrics services team use our platform, the Elluminate Clinical Data Cloud, to bring data together from any source, structure, or format and turn raw data into insights. We have a robust AI infrastructure and ML pipelines as part of this architecture. Across the clinical data lifecycle, there are many opportunities to apply AI/ML for greater efficiency and reduced cycle times, especially within the data integration and consumption components of the lifecycle where we are focused – from data transformation and mapping to data review and analytics.
Clinical data is a critical asset of life sciences, and companies face increased protocol and data complexity, adding pressure to cycle times. Adopting AI to eliminate inefficiencies and reduce manual work is imperative for companies to remain competitive and achieve these outcomes. Stepwise or incremental methods still work in clinical trials, but there is also an urgency to digitally transform and adopt modern infrastructures that make data available and accessible for efficient decision-making.
eClinical Solutions enables our clients with the foundational architecture that brings data together for advanced applications, along with embedded capabilities to apply AI/ML-driven automation. Examples are data mapping and classification, data review, and data insights. Within the data management workbench component of our overall platform, there is AI-enabled automation of data review. This enables clinical data management teams, including data managers and medical data reviewers, to be efficient while ensuring data quality at scale.
Every company in life sciences is on a journey to digitize, and the ability to leverage advanced analytic capabilities is a key goal within this transformation. Some use cases across the industry are more mature than others, but AI/ML has shifted from hype to reality in clinical development. Companies that do not explore and adopt AI will be left behind. Adoption may vary, but nearly all sponsors are incorporating AI into their strategies, and many technology providers in life sciences now provide substantial AI capabilities. Incorporating AI/ML into clinical data lifecycle processes is necessary to keep up with the evolving clinical trial landscape as trial complexity, data volume, and data variety continue to increase. AI carries enormous potential for our industry, however, identifying appropriate use cases is critical for life sciences organizations to see value and generate positive business outcomes.
How has the pandemic influenced eClinical Solutions’ approach to trial optimization and the overall R&D process?
Indupuri: The pandemic pushed our industry to break away from the status quo in many areas. Digitization and patient-centricity were already a priority pre-pandemic, but the resulting disruption forced stakeholders to adopt new approaches and utilize technology at a greater rate. There has been greater intent around digitization, and the pandemic accelerated the move toward digital transformation and decentralization. We’re seeing more industry-wide dedication in incorporating those lessons learned and collaborating around decentralized clinical trial (DCT) models. Many of these trends are positive for our space, but they increase the need for innovation and speed within our data management processes. The burdens of clinical data have grown as more technology systems, data types and data streams are incorporated, such as those from wearables, biomarkers, and labs. Even without these trends, the diversity of data was steadily growing pre-pandemic. As the modalities for DCT data acquisition expand, so do the data challenges.
The other factor is innovation and a recognition of the need to move science rapidly. Life sciences companies are seeing themselves more and more as data and technology companies. With developments in gene therapy and precision medicine over the last decade, scientific progress has been ground-breaking. Companies increasingly recognize that they must have a future-ready infrastructure to extract the full value of their data and be prepared for what’s next.
Since we offer biometrics services and have a team helping our clients with data aggregation, data review and data analytics, we have seen these industry developments unfold from the perspective of both technology foundations and from the vantage point of the teams aiming to unlock optimal value from their data.
Could you discuss eClinical Solutions’ approach to data integrity?
Indupuri: Data integrity is extremely critical in clinical trials, data management, and regulatory compliance. Our Elluminate Clinical Data Cloud foundation capabilities have been engineered with this in mind, and the platform ensures data integrity by embedded end-to-end automation from seamless ingestion to insights while providing transparency into the entire pipeline. All source or raw data ingested into Elluminate, irrespective of the size and complexity of the data, is retained. In addition, there are several built-in capabilities to validate data and metadata, including statistical methods and ML for detecting data issues. We also have embedded several workflows around data issues management, audit trail review, logging, etc., for monitoring and oversight. We have taken an approach to integrate all data products within the platform to minimize the need to make copies for different use cases and by providing complex features such as change data capture, blinding, masking, etc., to support all aspects of clinical data handling, enabling a software-driven approach to data integrity and quality.
What potential benefits and drawbacks does automation bring to the clinical data review process?
Indupuri: Today’s trial complexity means more data collected and more data to interrogate and clean. Detection of anomalous data is difficult and often a manually intensive, time-consuming effort. Automation can bring numerous benefits to the clinical data review process, from reducing tedious or manual tasks to surfacing critical errors and improving patient safety and efficacy. Automation augments the work of clinical data managers and reviewers so they can utilize their expertise to focus on areas of greatest impact. For sponsors and sites, automating data review will enable the delivery of high-quality data for analysis and submission in real time.
When applying AI and automation to data review processes, we need to ensure it’s a human-in-the-loop (HITL) approach. Because we have an in-house biometrics services team and know the domain and workflows, we see firsthand the value that AI-enabled clinical data review can deliver to sponsors. A single trial can now generate billions of data points. Traditional data management approaches no longer scale to ensure data integrity and quality efficiently in this data environment.
The potential benefits are many, but they require change management – a key consideration for achieving desired outcomes. Process changes, and upskilling are required so that users can operationalize these advanced capabilities. A key facet of adoption is also building trust in the models, and to do that, you must collaborate with stakeholders on their role as HITL and provide insight into the inner workings of the AI models and how they are arriving at their predictions. Explainability and interpretability are imperative for building the trust that drives adoption. For data review, there will still be humans to verify results and make decisions along the way – the stakes are too high not to when you are talking about patient data and safety.
How can AI/ML technologies assist in detecting anomalous data and identifying data quality issues?
Indupuri: Anomaly detection algorithms can identify data outliers that may indicate data quality or safety signal issues, such as adverse events vs. concomitant medications. It allows the data manager to have the models take a first pass at identifying the issues that need expert review. The ML models can also classify data points, freeing up the data manager’s time focus on review of the most critical data. Another example is labs or vitals. As a human, looking at trends is a difficult task; we can only take in two or three variables or dimensions of data at any point. If you take that and think about looking at thousands of values in a study, it’s hard to identify those inconsistent trends.
As the CEO of eClinical Solutions, how do you encourage an innovative mindset within your team? How do you choose what not to focus on?
Indupuri: It’s imperative to welcome the ideas and inputs of your team and maintain a focus on communication if you want your team to have an innovative mindset. No single leader has all the answers, and you need to model that approach so that leaders throughout the organization adopt it. If you want an innovation-led company, you need open doors and a culture where ideas are welcome and encouraged. We talk a lot about disagreement versus disapproval. You need to cultivate an environment where it’s safe to disagree. Disagreement does not mean disparaging others’ ideas but working together to uncover the best approach for success. When you are aligned on common goals that are well-communicated, and everyone has a voice, your team is more apt to be inspired and empowered about their impact. It starts with modeling that approach from the top down so that your broader team knows it’s ok to take risks. If your organization is not leading with that open mindset, creativity can be easily stifled.
Choosing what to focus on is about monitoring signals and revisiting goals regularly, using data to know when it’s time to pivot or shift attention. It’s about listening to your clients, partners and team and remaining flexible while using information to guide your focus against your larger vision and goals.
What role do you foresee for analytics and AI in clinical trials optimization and drug development in the months to come?
Indupuri: I am excited and optimistic about the possibilities of advanced analytics and machine learning. I think we will see continued exploration of large language models and generative AI in clinical trials. Every company in our space is looking at ways to pull all types of data together and make it available and accessible to solve complex problems faster. Tech and modern data infrastructure are now making this possible. A year from now and beyond, I see our industry tapping into the power of AI/ML for semi-structured or unstructured data, where it has the potential to be incredibly powerful. For example, ML could be used to analyze information from study protocols or literature and turn these into insights that aid in better research or improved protocol designs with high performance and at scale.