Senior Specialist in Applied Research – Advanced Technologies CoE
Fortunately, advances in biomedicine have coincided with technological progress, creating unprecedented opportunities to address clinical trial-related challenges through artificial intelligence (AI). AI is increasingly being integrated across the pharmaceutical development process from drug discovery to clinical trials.
During the trial planning stage, AI can analyze data from past studies and real-world sources to help research teams choose appropriate endpoints and anticipate potential operational challenges. By bringing together information from multiple systems, it becomes easier to identify risks early and reduce the likelihood of protocol amendments, which are often responsible for costly delays.
AI is also transforming patient recruitment. Using technologies such as natural language processing, researchers can review electronic health records and other data sources more efficiently to find individuals who match study requirements. This approach can significantly reduce the time and effort involved in manual screening. In addition, AI-powered chatbots and digital screening tools can help collect preliminary patient information and guide potentially eligible candidates through the next steps in the recruitment process.
Beyond recruitment, AI accelerates data analysis throughout the study. Machine learning models can detect emerging safety concerns, uncover meaningful biomarkers, and streamline data management tasks such as data validation and cleaning. As a result, research teams can generate insights faster and make more informed decisions throughout a clinical trial. In the following sections we will explore a few examples of how AI is transforming how clinical trials are conducted.
One of the most significant bottlenecks in clinical trials is the recruitment phase, with 80% of trials experiencing delays and 37% of sites failing to recruit a single participant. Overly restrictive eligibility criteria may be one reason for this. While these criteria help protect patients and ensure reliable results, they can also make it harder to recruit enough participants.
As a result, identifying patients who meet the eligibility criteria can significantly delay trials. Another common challenge in clinical trials is patient dropout, which can introduce bias and compromise data analysis.
AI can improve these processes by using data to broaden eligibility criteria without increasing risk. It can also speed up the initial screening process and help sponsors identify patients who may qualify for a clinical trial. This streamlines the recruitment process, allowing teams to focus on engaging suitable candidates sooner. Moreover, AI could proactively identify patients at risk of dropping out and enable research teams to provide personalized support to improve adherence and retention throughout the trial.
Improved matching between patients and studies can also increase the likelihood that participants are enrolled in trials that are most relevant to their specific health conditions and needs, thereby increasing their engagement and reducing the likelihood of dropping out.
According to the E6 (R3) revision, participant consent is now viewed as an ongoing process that must be updated whenever new information becomes available.While the revision advocates for the use of technology, as we have mentioned earlier, it goes further by demanding that systems such as digital health technologies (DHTs) and electronic source data (eSources) be validated and checked to ensure they are not just secure and auditable but also fit for purpose based on the level of risk and data criticality. Regardless of the medium you use to capture, store, or manage trial data, the new GCP guidelines’ media-neutral stance means GCP principles will be applied consistently across all technological platforms to ensure the same ethical and scientific standards are upheld, whether you collect data through traditional paper methods or advanced digital systems.
Although our primary focus so far has largely been on the technical changes being ushered in by E6 (R3), the recognition of the different roles and responsibilities Service Providers may take on beyond the traditionally assigned ones in the ICH E6 (R3) update is a significant change worth mentioning. However, it is important to remember that, while a sponsor may delegate roles and activities during the trial, they hold full accountability for the quality and integrity of the entire trial and not the delegated parties. As we highlighted earlier, investigators have more roles to play in the trial and within their new responsibilities, R3 also gives them the authority to delegate to service providers as well, in collaboration with a sponsor who can also support them by identifying appropriate providers for the trial.
AI is also transforming data management by automating many of the traditionally labor-intensive processes required to ensure high-quality research data.
Natural language processing (NLP) algorithms play a crucial role in managing unstructured clinical information. By automatically extracting and structuring data from study protocols, clinical notes, and source documents, these algorithms can streamline the creation of electronic case report forms (eCRFs) and automate data entry, reducing study setup time, administrative work, and transcription errors.
Once the eCRFs have been created and the data entered, the next step is data cleaning. This is often one of the most time-consuming parts of a clinical trial. AI can help by detecting inconsistencies, missing values, and errors across large datasets. NLP algorithms can help standardize free-text entries, while machine learning models can identify anomalies and predict missing data based on patient characteristics and study context. AI can also spot unusual physiological values or missing data and generate queries for someone to review. Compared with traditional methods, this may cut the time needed for data cleaning by 60–80%.
Clinical trial data are usually analyzed through structured workflows built and managed by biostatisticians. AI can support these workflows by helping researchers gain meaningful insights much faster than with traditional methods. By processing large volumes of data in real time, machine learning models can support exploratory analyses early in the study lifecycle, allowing research teams to identify relevant patterns and make more informed decisions throughout the trial, instead of waiting for scheduled interim analyses.
This capability is particularly valuable in studies that rely on connected devices to collect continuous biometric measurements. As thousands of data points are transmitted to study databases each day, AI systems can rapidly aggregate and analyze the information, detecting emerging trends and changes in patient responses almost in real time. This continuous monitoring approach provides investigators with a more up-to-date understanding of the study’s progress and participant outcomes.
AI also enables the development of dynamic risk prediction models that improve as more trial data are collected. By integrating diverse sources of information, including clinical assessments, laboratory results, imaging data, and digital biomarkers, these models can generate real-time predictions of individual patient outcomes as well as overall trial performance. Compared with conventional statistical methods, which typically evaluate a limited number of variables, AI algorithms can simultaneously analyze large, complex datasets, leading to improved predictive accuracy.
In addition, AI strengthens patient safety monitoring by evaluating multiple data streams in parallel. Adverse event reports, laboratory findings, and clinical measurements can be assessed together, allowing advanced pattern-recognition algorithms to identify patterns or correlations and detect potential safety signals at an early stage. As a result, researchers can respond more quickly to emerging risks and improve both trial oversight and participant safety.
Although the potential is clear, this progress also comes with challenges.
First, AI models require large amounts of high-quality data for training and validation. However, clinical trial data are often fragmented across different systems, collected using different standards, and subject to privacy and access restrictions. The lack of harmonised and representative datasets remains one of the main barriers to developing reliable AI solutions. Efforts to improve data standardization and interoperability will therefore be essential.
A second challenge is the highly regulated nature of the pharmaceutical industry. Regulatory requirements such as the EU AI Act, FDA guidance, and GxP requirements impose strict expectations regarding data governance, risk management, transparency, and validation. Any AI tool used in a GxP environment must demonstrate that it is fit for purpose, reliable, and compliant with applicable regulations before being adopted in practice.
Finally, reliability and transparency remain key concerns. Many advanced AI models operate as “black boxes,” making it difficult to understand how outputs are generated. For this reason, a human-in-the-loop approach remains crucial. AI should support, rather than replace, clinical, statistical, and regulatory expertise, ensuring that critical decisions continue to benefit from human judgement and oversight. Regulatory agencies and organizations such as the WHO have emphasized the importance of transparency, accountability, and continuous monitoring of AI systems.
Despite these challenges, the future of AI in clinical trials is promising. As data standards improve and regulatory frameworks mature, AI is expected to enhance efficiency, accelerate evidence generation, and support more informed decision-making throughout the clinical development process. For sponsors and CROs, however, the real differentiator will not be the technology itself but the ability to embed it into validated, auditable processes in which human accountability remains clearly defined.
In our experience, the main barrier to the effective adoption of AI in clinical trials is rarely the performance of the algorithm alone. More often, it lies in the readiness of the processes, data, and governance framework surrounding it. Data that have not been standardized during the protocol and eCRF design stages are difficult to use reliably later in the trial, and an AI-enabled tool introduced without a clear intended use, appropriate risk assessment, and proportionate qualification is difficult to justify in a regulated environment. Outputs also need clearly defined ownership: without established review and escalation responsibilities, they risk being either disregarded or relied upon without sufficient scrutiny.
A pragmatic approach is therefore to introduce AI progressively, starting with clearly defined use cases in which its performance can be measured and reviewed, and any issues can be corrected. High-volume activities such as data review support, query management, medical coding assistance, deviation triage, and the identification of site-level risk signals within a risk-based quality management framework may represent suitable starting points. Applications that directly influence patient safety, efficacy assessment, or critical trial decisions require a higher level of validation, explainability, and human oversight.
Several factors support this transition: data standards and quality requirements defined early in study design; multidisciplinary collaboration between clinical operations, data management, biostatistics, quality, regulatory, and technology functions; risk-proportionate validation and lifecycle management of computerized systems; and training for the professionals who review and act upon AI-generated outputs.
For companies beginning this transition, the objective is not to adopt AI as broadly or rapidly as possible but to match specific operational needs with solutions that remain transparent, auditable, and under human control. The organizations we see progressing fastest start small, evaluate results objectively, and scale only what has proved reliable. Introduced in this way, AI moves beyond technological promise and becomes a practical capability that supports more efficient and reliable clinical trial conduct.
-1.webp)
