AI-powered clinical trial systems still fragmented

The review found that while AI is making inroads in specific use cases, such as computer vision for medical imaging, multimodal AI for integrated datasets, federated learning for privacy-preserving analysis, predictive inference for outcome forecasting, and digital twins for trial simulations, these capabilities are unevenly distributed. Large language models (LLMs), explainable AI (XAI), and machine learning operations (MLOps) are only sporadically implemented and typically restricted to niche functions.

A new systematic literature review published in Information warns that while artificial intelligence (AI) and cloud-native computing are reshaping clinical research information systems, the industry is far from realizing their full potential. The research, conducted by Isabel Bejerano-Blázquez and Miguel Familiar-Cabero, finds that vendor offerings remain fragmented, with incomplete integration of critical medical features across the clinical trial lifecycle.

Titled “On the Application of Artificial Intelligence and Cloud-Native Computing to Clinical Research Information Systems: A Systematic Literature Review”, the study analyzes 7,283 records and distills findings from 181 relevant works. It offers a structured, 360-degree blueprint for evaluating AI-enabled clinical trial management systems (CTMS) and calls for higher alignment with Good Clinical Practice (GCP) standards.

Mapping the current landscape of AI and cloud in clinical trials

The authors examined the state of AI and cloud-native adoption within clinical research information systems (CRIS), focusing particularly on CTMS platforms that support study design, initiation, management, data handling, analysis, reporting, and regulatory submissions. Their findings reveal a market populated by technologically advanced yet siloed solutions, often strong in individual domains but rarely delivering complete end-to-end functionality.

The review found that while AI is making inroads in specific use cases, such as computer vision for medical imaging, multimodal AI for integrated datasets, federated learning for privacy-preserving analysis, predictive inference for outcome forecasting, and digital twins for trial simulations, these capabilities are unevenly distributed. Large language models (LLMs), explainable AI (XAI), and machine learning operations (MLOps) are only sporadically implemented and typically restricted to niche functions.

From a deployment standpoint, most leading solutions are delivered as cloud-based software-as-a-service (SaaS) hosted on major hyperscalers like Amazon Web Services, Microsoft Azure, or Google Cloud. This model provides scalability, flexibility, and cost efficiency, but does not automatically guarantee feature completeness or regulatory alignment.

Blueprint for a complete, GCP-aligned CRIS-CTMS

To address the gaps, the study introduces a 360-degree maturity framework that outlines the essential medical features a robust, GCP-compliant CRIS-CTMS should support. The blueprint spans the entire clinical trial lifecycle:

• Study Design: protocol planning, feasibility analysis, and resource allocation
• Initiation and Setup: site selection, investigator engagement, and ethics approvals
• Trial Management: recruitment, scheduling, and operational oversight
• Data Management: secure collection, integration, cleaning, and validation
• Analysis and Reporting: statistical modeling, safety reporting, and interim analyses
• Filings and Submissions: preparation of regulatory documents and dossiers
• Supply Management: inventory tracking, logistics, and distribution controls

The maturity model also includes a GCP alignment index to ensure that compliance and auditability are embedded in system design. This is critical for sponsors, contract research organizations (CROs), and regulators who must validate the integrity of trial data and processes.

When applied to a set of prominent vendors, including Saama Technologies, ConcertAI, Owkin, PathAI, Lantern Pharma, Unlearn, and AiCure, the blueprint revealed clear differences in maturity. Saama Technologies scored highest overall, followed by ConcertAI and Owkin, yet none demonstrated full integration across all domains. In every case, critical medical features were missing, often leading to operational silos and inefficiencies.

Strategic imperatives for vendors, sponsors, and researchers

For vendors, the priority is to bridge the gaps between isolated capabilities, ensuring that AI and cloud-native components are fully integrated across all trial domains. This means moving beyond specialized tools toward unified platforms that can handle every stage from protocol drafting to regulatory submission.

For sponsors and CROs, the blueprint offers a practical tool for procurement decisions. By mapping prospective solutions against the 360-degree framework, buyers can identify weaknesses before deployment and avoid costly retrofits or parallel systems. Given the uneven adoption of advanced AI techniques such as LLMs, XAI, and MLOps, procurement teams must also weigh the trade-offs between cutting-edge features and proven operational readiness.

Researchers are encouraged to focus on end-to-end assessments of AI and cloud-native impacts in clinical trial systems, rather than isolated technical demonstrations. The review highlights the need for empirical studies that measure not just algorithmic performance but also operational efficiency, compliance adherence, and long-term scalability in live trial environments. The authors also stress that ethical and regulatory considerations must remain central.