On the Application of Artificial Intelligence and Cloud-Native Computing to Clinical Research Information Systems: A Systematic Literature Review

Abstract

The pharmaceutical and biotechnology sector is an intricate and rapidly evolving industry encompassing the full lifecycle of drugs, medicines, and clinical devices. Its growth is driven by factors such as the aging population, the rise in chronic diseases, and the increasing focus on personalized medicine. Nevertheless, it also faces significant challenges due to rising costs, increased complexity, and regulatory hurdles. Through a systematic literature review (SLR) as a research method combined with a comprehensive market analysis, this paper explores how several leading early-adopter healthcare companies are increasing their investments in computer-based clinical research information systems (CRISs) to sustain productivity, particularly through the adoption of artificial intelligence (AI) and cloud-native computing. As an extension of this research, a novel 360-degree reference blueprint is proposed for the domain analysis of medical features within AI-powered CRIS applications. This theoretical framework specifically targets clinical trial management systems (CRIS-CTMSs). Additionally, a detailed review is presented of the leading commercial solutions, assessing their portfolios and business maturity, while highlighting major open innovation collaborations with prominent pharmaceutical and biotechnology companies.

1. Introduction

The global pharmaceutical and biotechnology industry is a complex and ever-evolving sector that encompasses the research, development, production, manufacturing, marketing, and distribution of drugs and medicines [1]. To ensure consumer safety, effectiveness, and drug quality, this industry is highly regulated [2]. Despite substantial investments in research and development, the sector faces several significant challenges [3].

Several factors drive the growth of the pharmaceutical and biotechnology industry. First, the aging global population increases the demand for treatments for age-related diseases such as cancer, heart disease, and Alzheimer’s [4,5]. Second, chronic diseases are the leading cause of death and disability worldwide, and this is driving the demand for new and innovative treatments and therapies [6,7]. Third, with the increasing focus on personalized medicine, tailoring treatments to individual patient needs enables the development of more effective drugs for specific populations [8,9].

The industry currently faces several primary headwinds and challenges. One major concern is the rising cost of drug development, which is placing significant pressure on healthcare companies to manage and control their expenses [10,11]. Additionally, heightened market competition has triggered price wars, subsequently reducing profit margins across the sector [12,13]. Compounding these issues are stringent and evolving clinical and data regulations, which often result in regulatory hurdles. These challenges can make the process of introducing new drugs and treatments to the healthcare market increasingly lengthy, complex, and expensive [14,15].

This trend can also be observed in terms of the impact on the return on investment in research and development (R&D). Global pharmaceutical research and development (R&D) spending increased from approximately 194 billion United States dollars (USD) in 2019 to USD 301 billion in 2023 [16], yet the return on investment has not shown a proportional improvement. However, despite this increase, nearly half of the top pharmaceutical multinational enterprises recorded negative R&D productivity [17]. This decline is particularly relevant to the clinical trial segment of the healthcare industry. In this area, several additional challenges to trial execution have been identified. These include high operational costs [18,19] and extended development timelines caused by patient recruitment and retention difficulties [20,21]. The industry also faces rising complexity in trial protocols and regulatory requirements [22,23,24,25,26], as well as persistent concerns regarding data quality and completeness [27,28]. To address the aforementioned productivity challenges, healthcare organizations have significantly increased their investments in digitalization across various business functions in recent years, especially in AI and cloud-native computing. These strategic investments aim to increase operational efficiency and sustain productivity levels [29,30,31].

In this context, the quantifiable benefits of AI adoption for clinical research optimization can be extracted from the literature review. Lam et al. [32] concluded that AI-assisted interventions perform better than standard care does, with improvements in 70% of the studies analyzed. Kavalci and Hartshorn [33] concluded that integrating eligibility criteria and ML-supported disease categorization features into study characteristics improved early clinical trial termination prediction. The model demonstrated good performance, achieving 70% overall accuracy and 42% balancing precision and recall (F1-score). Askin et al. [21] evaluated how clinical trial recruitment can be improved with ML, and how the application of the proposed model reduced the number of patients deemed unsuitable during chart reviews by 40.5% at the tertiary care center and by 57.0% at the community hospital. Additionally, Cascini et al. [34] explored the impact of AI in clinical trials. One of their key findings was that, when appropriate AI-assisted models were used, patient enrollment in breast cancer studies improved by 80% per month, resulting in an overall eligibility rate of 87.6%.

The preceding highlights how AI can improve the efficiency of clinical trials by enhancing performance and accuracy, while also improving patient recruitment and ensuring better alignment with eligibility criteria. Recent works, such as those by Ishii-Rousseau et al. [35] and Askin et al. [21], suggest a relationship between advanced computing and improved clinical outcomes. Several quantitative studies supporting this statement can be found in the literature, as outlined below. On the one hand, Van Leeuwen et al. [36] studied the application of intracranial large vessel occlusions in stroke in randomized trials, considering AI costs and performance. They reported a total cost saving of USD 11 million for each yearly cohort of patients in the country under study. On the other hand, Tsiachristas et al. [37] investigated the application of AI to quantify coronary computed tomography angiography. They observed that applied AI-based risk classification in routine coronary computed tomography implementation delivered an incremental cost-effectiveness ratio of USD 1693 to USD 4008. Nevertheless, while the literature on AI cost-effectiveness is promising, it remains limited and requires further evidence and case-by-case evaluations [38,39,40]. This limitation reflects the current immaturity of AI adoption in clinical research, as highlighted by the findings presented here. However, with increasing digitalization among pharmaceutical and biotechnology companies, AI integration is expected to accelerate.

In summary, the pharmaceutical and biotechnology sector is driven by factors such as an aging population, the increase in chronic diseases, and a growing focus on personalized medicine. This industry also faces significant challenges, including rising costs, increased complexity, and regulatory hurdles. In response, early adopter healthcare companies are making substantial investments in CRISs, particularly leveraging AI and cloud-native computing to drive productivity and support sustainable growth. Although some studies show a causal link between advanced computing capabilities and better clinical research outcomes, additional empirical evidence is needed as AI adoption in healthcare expands. In this vein, this paper aims to provide a deep understanding of the process by which leading pharmaceutical and biotechnology organizations are adopting AI and cloud-native computing. Our study explores the challenges influencing the adoption of emerging technologies in clinical research. It also analyzes emerging business and research opportunities from integrating these technologies into computer-based CRIS-CTMS solutions. The findings contribute to business acumen, theoretical development, and practical applications in the sector, filling a gap in the current literature.

The remainder of this paper is structured as follows. Section 2 justifies the theoretical foundation of this work, positioning our contribution within established models in healthcare informatics. Section 3 provides a detailed account of the methodology adopted to analyze prevailing trends and strategic investments that are shaping the future trajectory of the healthcare industry. Section 4 presents the results arising from the SLR. This includes an overview of market research and business leadership, an assessment of the maturity associated with the digital transformation of CRIS-CTMS solutions, and a review of the main discussion points and contributions of our work. Finally, Section 5 concludes the paper by summarizing the principal findings, reviewing the main contributions and limitations, and outlining future research.

2. Theoretical Foundation

To position our contribution within the broader theoretical landscape, this section surveys the related literature on established models in healthcare informatics. For this purpose, existing work in the following four domains is examined: healthcare market and clinical business research; AI and cloud-native technologies in healthcare; CRIS-CTMS-related literature; and AI applied to medical features in CRIS-CTMS. The rationale for selecting these areas is grounded in the motivation and interdisciplinary nature of our work, as previously outlined.

In exploring the intersection of the healthcare market and clinical research, several studies offer valuable insights into current challenges and emerging technologies. Schuhmacher et al. [17] analyze the pharmaceutical industry over the period from 2001 to 2020, focusing on productivity and R&D expenditures. While this study provides a comprehensive examination of market dynamics, it does not address the potential benefits of AI in helping these companies overcome current challenges. Bentley [18] and Fogel [19] highlight that financial constraints continue to present a substantial barrier, with more than 20 percent of failed phase III trials attributed to budgetary limitations and the average cost per patient reaching approximately USD 42,000. In addition to financial obstacles, prolonged development timelines, often extending up to 15 years, pose further difficulties, largely due to persistent issues in patient recruitment and retention [20,21]. Additionally, Schuhmacher [41] expands on the broader digital transformation in the sector and AI’s potential for value creation processes in pharmaceutical and biotechnology companies. However, these prior works do not examine AI adoption maturity among industry leaders, nor do they consider the requirements of CRIS-CTMS solutions. Fernald et al. [3] analyze the productivity gap within the pharmaceutical industry, on the basis of data from 200 companies and more than 80,000 clinical trials spanning from 2012 to 2023. Nonetheless, this study overlooks AI-driven strategic roadmaps or actions taken by industry leaders to address these challenges through digitalization. Furthering this line of inquiry, Wu et al. [42] examine R&D productivity in China, alongside the potential of AI technologies to enhance new drug development. However, this study lacks global benchmarking of AI adoption across the industry and does not address CRIS-CTMS transformations within the companies under consideration.

To contextualize AI within the broader healthcare technology landscape, other research explores its applications in clinical decision-making and operational processes. Koo et al. [43] explore AI’s impact on healthcare, particularly in nursing care, emphasizing its potential to enhance clinical decision-making, patient care, and operational efficiency. Similarly, Clark et al. [44] highlight AI’s potential to revolutionize healthcare by improving patient safety, diagnostic accuracy, and treatment efficacy while reducing costs. Their systematic review identified key areas of interest, including medical imaging, drug discovery, wearable devices, and ethical AI integration. However, their survey lacks insight into how leading organizations are advancing in these areas. In the same vein, Jiao et al. [39] assess the economic impact of AI in healthcare and evaluate its overall economic value, though this study does not address the specific actions undertaken by healthcare organizations. Ali et al. [45] investigate the role of AI in improving diagnosis and treatment, identifying key functionalities, challenges, and future research directions. Kolasa et al. [46] offer a systematic review of 220 studies focused on ML applications in healthcare, highlighting its primary uses in clinical prediction and disease prognosis. Despite their contributions, none of these studies provide a comprehensive analysis of AI and cloud-native technologies for developing clinical systems or offer guidance for assessing AI adoption maturity.

Further studies have attempted to explore the evolving CRIS-CTMS landscape. Baskerville et al. [47] propose a framework for clinical system development, focusing on the following four components: context, desired outcome, intervention, and effective outcome. However, it does not address the medical or digital capabilities of clinical systems. Madhavan et al. [48] examine a national clinical information system in Wales, identifying its unique features, through semi-structured interviews, a quantitative survey of 559 clinical users, and a focus group. Nevertheless, this market-specific study lacks insights into how AI and cloud-native technologies could enhance those features. Combi and Pozzi [49] conduct a literature review on clinical information systems and the application of AI techniques. They offer a taxonomy and research directions, including data analytics, data mining, and natural language processing. However, the article does not assess the impact or adoption of these technologies, nor does it analyze the commercial provider landscape. A more recent contribution to the field is made by Kapustina et al. [50]. While the authors describe AI integration in drug development and production, the study lacks market justification. Additionally, it does not offer a maturity analysis of AI adoption by leading companies in the sector.

Research focused on the medical features relevant to CRIS-CTMS contributes additional depth to the discourse and helps to frame our comprehensive, 360-degree blueprint for maturity analysis. McCaskell et al. [51] describe how key medical activities in the CYCLE pilot trial were organized across domains such as study initiation, execution, and monitoring. Singh et al. [52] emphasize the importance of study design to ensure ethical standards and scientific integrity. This domain is crucial for accurately assessing the safety and efficacy of prototype, modified, and next-generation vaccines. Kandi and Vadakedath [53] focus on data analysis and regulatory compliance in clinical research, underscoring their role in ensuring the accuracy, reliability, and ethical conduct of trials. Rombach et al. [54] stress the need for clear statistical analysis and transparent reporting. Their findings revealed that many randomized clinical trials failed to report treatment effect measures, adequately address missing data, or conduct necessary sensitivity analyses, which called for improved reporting practices. Copland et al. [55] identify the critical role of supply management in clinical trials to ensure the timely and accurate distribution of investigational equipment and materials to participating sites. Iacoviello et al. [56] demonstrate how interpretable ML models can accurately stratify mortality and hospitalization risk in heart failure patients with thyroid dysfunction, offering a novel tool for personalized medicine. This approach to risk stratification not only supports individualized clinical decision-making, but also enhances the organization of care pathways, particularly for patients with comorbidities. While the reviewed research provides valuable insights into the medical feature domains required for advanced AI and cloud-native CRIS-CTMS, none offer a comprehensive view of the capabilities needed or a blueprint for assessing adoption and guiding future developments.

Collectively, the existing body of literature demonstrates growing interest in AI’s potential within the pharmaceutical and healthcare sectors. Nonetheless, despite this momentum, a systematic evaluation of AI adoption among leading industry stakeholders remains absent. Moreover, current studies fall short of analyzing strategic roadmaps or the organizational transformations required for effective CRIS-CTMS integration. Similarly, research on AI and cloud-native technologies in healthcare has focused primarily on isolated therapeutic areas and lacks a cohesive blueprint to evaluate AI adoption maturity. The literature on medical features for advanced CRIS-CTMS solutions has identified key functionalities and challenges, but these insights remain scattered, without a unified blueprint for assessing digitalization in clinical trial management systems and guiding future advancements. Addressing these gaps in a seamless manner is crucial for fostering an integrated approach to AI and cloud-native adoption in clinical research information systems, enhancing innovation, efficiency, and patient outcomes.

3. Methodology

This section outlines the methodological framework employed to examine the adoption of AI and cloud-native technologies in clinical research information systems. The methodology begins with a discussion of the motivation and the rationale for the selected research method. It then introduces the formulation of research questions and theoretical propositions that provide structure to guide the analysis. The process continues with the definition of inclusion and exclusion criteria, the identification of relevant data sources, and the design of a comprehensive search strategy. Finally, it describes the procedures used for structured data extraction. Each of these steps is outlined in the subsections that follow.

3.1. Motivation and Research Method

In a workshop held in Krakow, Poland, as part of the Annual Conference of the Pharmaceutical Contract Management Group in September 2022, more than 200 experts from around the world argued that by 2050, professionals working in clinical research will likely be data scientists [57]. This bold statement resonates with the vibrant transformation process in which the healthcare industry is currently immersed.

This paper aims to gain profound insights into the maturity journey toward the adoption of AI and cloud-native computing by some of the top, early adopter pharmaceutical and biotechnology companies. The main motivation of this work is to explore the challenges driving the adoption of these novel technologies in clinical research, as well as the business and research opportunities they present for computer-based CRIS–CTMS solutions. For that purpose, leading providers in this field are reviewed to investigate how holistic their approaches are. Additionally, their most representative open innovation and partnership initiatives with commercial pharmaceutical and biotechnology companies are examined to devise future lines of research.

Owing to the high number of disparate and heterogeneous studies, conducting a rigorous and reproducible analysis that ensures analytical credibility in the gathered findings requires relying on a systematic method of literature analysis [58,59,60]. The meta-analysis methodology was initially considered due to the large volume of the related literature. However, the empirical data in this field still lack strong adoption, maturity, and statistical homogeneity. Accordingly, a SLR may be a more appropriate approach for synthesizing research in this domain, as meta-analysis benefits from studies with greater similarity for effective analysis and synthesis [61,62,63].

Therefore, the analysis in this paper is based on a SLR to collect, identify, evaluate and analyze all relevant literature on the topic under study. This methodological approach enables the addressing of a specific research question in a transparent and reproducible manner, while critically assessing the quality of the evidence [64,65]. To further strengthen our research in terms of richness, reproducibility, trustworthiness, and utility, we incorporate the principles proposed by Simsek et al. [66]. These principles are as follows: transparency, coverage, saturation, connectedness, universalism, and coherence.

Transparency is reflected in how clearly the methodological procedures are defined and documented. The three main components of transparency are as follows: clearly stating the reviewer’s process and methods; assumptions underlying the review should be explicitly outlined; and distinguishing literature observations from findings. Regarding process and methods, this SLR follows the methodology outlined by Brereton [67] and Kitchenham et al. [68], which is structured into planning, conducting, and results. The standardized data extraction spreadsheet used in the process is provided in Appendix A. The assumptions underpinning the review are explicitly stated through research questions (RQs) and theoretical propositions (TPs) in Section 3.2. These elements guide the analysis in an open, transparent, and reproducible manner. Additionally, a clear distinction is made between literature observations and findings. Section 2 highlights the existing gap in the literature. The findings of our study, clearly delineated in Section 4.5, are linked to the TPs to facilitate a rigorous validation of the initial hypotheses.

The principle of completeness underscores the importance of achieving thorough coverage of all the relevant literature. This encompasses both traditional scientific databases and gray literature. To meet this principle, our SLR surveys 181 references, covering factors influencing growth in the pharmaceutical and biotechnology sector; industry headwinds and challenges in executing clinical trials; quantifiable benefits of AI adoption; and efficiency improvements in clinical research. Additionally, it examines the overall pharmaceutical market outlook regarding AI investments and digitalization maturity, the theoretical positioning of computer-based clinical information systems, and the specific actions taken by healthcare industry leaders in terms of readiness and maturity for transforming their clinical systems. Furthermore, the review identifies gaps in the literature concerning theoretical models defining medical features for clinical information systems, explores the required capabilities for leading suppliers of clinical information systems, and discusses future research directions. The inclusion of gray literature in our SLR is further justified in Section 3.4, where the applied methodology for assessing gray literature is outlined.

Saturation indicates that no new data or insights emerge that would further contribute to the development of a category or framework, signifying that all the relevant literature has been thoroughly explored. This principle is supported by several strategies in our SLR. First, three search strings with Boolean operators (see Section 3.3) are used to maximize the collection of published documents. Second, we incorporate multiple peer-reviewed article databases, each specializing in different knowledge domains, ensuring broad and comprehensive literature coverage (see Section 3.4). Finally, the inclusion of the assessed gray literature provides insights that are often overlooked in academic publications, thereby enhancing the depth and thoroughness of the review.

A fundamental aspect of the review is its attempt to identify and emphasize connections across the body of literature surveyed. As highlighted in Section 2, the existing research lacks integrated studies on AI and cloud-native technologies in healthcare, focusing instead on isolated therapeutic areas. Additionally, while the literature on medical features for advanced CRIS-CTMS solutions highlights key functionalities and challenges, it remains fragmented, lacking a unified approach for evaluating digitalization in clinical trial management systems. A key contribution of the present work is addressing these gaps in a seamless and interconnected manner. To achieve this, our SLR links the pharmaceutical and biotechnology sector with AI technologies, aligning them with business and clinical system requirements. This connection has led to the development of an integrated reference blueprint for implementation and benchmarking, which is further discussed in Section 4.3.

To ensure objectivity, the principle of universalism is maintained throughout the review process. To fulfill this principle, this SLR establishes clear inclusion and exclusion criteria (see Section 3.5) to ensure that appropriate studies are considered while avoiding bias. Additionally, a systematic and transparently documented search protocol was implemented. As described in Section 3.6, two independent reviewers assessed the literature, resolving disagreements through discussion to reach a consensus. This collaborative approach minimizes individual bias and ensures that conclusions are based on a balanced interpretation of the evidence.

Lastly, systematicity emphasizes that explicit methods are consistently applied throughout the review process for coherence. Our research design adheres to the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) methodology [69]. It provides a structured framework for conducting systematic reviews and meta-analyses, ensuring consistency and reproducibility. To further reinforce this principle, we structure our SLR according to the systematic procedure described by Brereton [67] and Kitchenham et al. [68]. This approach is organized into the following three steps: “planning”, which is detailed in Section 3.2, Section 3.3, Section 3.4 and Section 3.5, defining the scope and methodology; “conducting”, which is explained in Section 3.6 and Section 3.7, where the literature search, selection, and data extraction processes are described; and “results”, which is covered in Section 4, where the synthesis and presentation of findings occur. This approach ensures that all review practices are applied rigorously and coherently.

3.2. Research Questions and Theoretical Propositions

This study addresses the following RQs. Related TPs are defined for each research question to guide the analysis throughout the research, enabling a more effective discussion of the findings in Section 4.5.

• RQ1. What are the main growth drivers and challenges in the healthcare industry?

• TP1: Healthcare industry growth is driven primarily by increasing global healthcare needs, particularly those related to chronic and age-related diseases, and sector digitalization.
• TP2: Key challenges include rising development costs, process and regulatory complexity, and inefficiencies in clinical trials and drug development.

• RQ2. What opportunities does the adoption of AI and cloud-native computing bring to clinical research, to address the headwinds identified in RQ1?

• TP3: AI and cloud-native computing present opportunities to overcome key challenges in clinical trial research, such as enhancing patient recruitment and ensuring regulatory compliance.
• TP4: These technologies can significantly accelerate the drug development process, reducing both time and cost while improving treatment accuracy.

• RQ3. To what degree of maturity are the main pharmaceutical and biotechnology companies adopting those technologies in their CRIS-CTMS solutions?

• TP5: AI and cloud-native computing adoption in CRIS-CTMS solutions remains at an early stage for most pharmaceutical and biotechnology companies.
• TP6: Adoption levels tend to correlate with company size and market position, with larger firms integrating these technologies at a higher rate than smaller organizations do.

• RQ4. What are the key feature domains and medical informatics capabilities that AI-powered CRIS-CTMS applications should implement to deliver holistic, advanced clinical research analytics?

• TP7: There is a gap in the literature regarding a comprehensive view of these features in a unified reference blueprint. This is crucial for benchmarking solutions and guiding future research.
• TP8: AI-powered CRIS-CTMS applications should prioritize seamless integration of medical features across all trial domains.

• RQ5. What portfolios, open innovation initiatives, and degree of completeness do leading AI-powered CRIS-CTMS providers have regarding the key feature domains identified in RQ4?

• TP9: Leading AI-powered CRIS-CTMS providers leverage AI and cloud-native technologies, which are often developed in collaboration with pharmaceutical companies.
• TP10: Medical capabilities among leading AI-powered CRIS-CTMS providers remain fragmented, focusing on specific medical domains.

3.3. Search Strings

The search terms used in the SLR were constructed based on expert domain knowledge of the subject. To construct search strings (SSs) from selected keywords, we applied the Boolean operators “OR” and “AND”. Consequently, several search strings have been developed to maximize the retrieval of published documents and ensure comprehensive coverage of the relevant literature [70]. The following three search strings were used to scan the data sources:

• SS1: ((“pharmaceutical” OR “healthcare” OR “clinical” OR “biotech”) AND (“AI” OR “ML” OR “cloud”) AND (“clinical information system” OR “CTMS”))
• SS2: (“Amazon” OR “Google” OR “Azure”) AND “pharmaceutical” AND (“Roche” OR “Janssen” OR “AstraZeneca” OR “Pfizer” OR “GlaxoSmithKline” OR “Bristol Myers Squibb” OR “Moderna” OR “Merck” OR “AbbVie” OR “Regeneron” OR “Gilead Sciences”)
• SS3: (“clinical research” AND (“Saama” OR “Owkin” OR “ConcertAI” OR “PathAI” OR “Lantern Pharma” OR “Unlearn” OR “AiCure”))

3.4. Selection of Data Sources

Based on the knowledge domains reviewed in this work (i.e., engineering, information systems, clinical research, healthcare, AI, and cloud computing), the following peer-reviewed article databases were selected: ACM Digital Library, IEEE Xplore, SpringerLink, ScienceDirect, Oxford Academic, PubMed, and JMIR. Additionally, and based on previous research experience on the subject, the authors added the following peer-reviewed article publishers: MDPI and Sage Journals.

In addition, gray literature limited to the form of corporate information, article preprints, press releases, and market reports was also considered via hand searching. This type of literature can generate important contributions to a SLR, providing insights that are sometimes missed in academic publications, improving the comprehensiveness of reviews and fostering a balanced picture of the gathered evidence. This contributes to building a bridge to the external world grounding academic research in practical and industrial contributions [71,72,73,74,75,76,77]. In this context, non-peer-reviewed records were evaluated using the AACODS framework [78], a widely recognized method in gray literature reviews [79,80]. This checklist assesses the concepts of authority, accuracy, coverage, objectivity, date, and significance within the context of gray literature. Included records were searchable, accessible, and clearly attributed, with no bias identified. They were linked to other materials and demonstrated comprehensive coverage of the topic, including recent findings (within the last five years).

3.5. Definition of Inclusion and Exclusion Criteria

Prior to determining the search strategy and data extraction, the SLR must define the criteria to ensure that appropriate studies are considered and bias is avoided. The inclusion and exclusion criteria in our research were based on the parameters defined by De Luzi et al. [81], Khan et al. [82], and Palmer et al. [83].

Four criteria for study inclusion were defined. Each individual inclusion criterion (IC) is listed as follows:

IC1: Language: The literature must be written in English only.

IC2: Accessibility: The literature is electronically available, either in a peer-reviewed article database or via Internet search.

IC3: Content Scope: The search terms have to be used in the context of the reported outcomes.

IC4: Relevance to research questions: The literature focused on the use of AI and cloud-native technologies and/or clinical research and clinical research information systems.

Four criteria for study exclusion were defined. Each individual exclusion criterion (EC) is listed as follows:

EC1: Not in the inclusion criteria: Literature not fitting in any of the IC1 to IC4 is excluded.

EC2: Fragmented results: Studies with findings reported across multiple publications are removed.

EC3: Duplicate literature: Exact or semantically identical studies are not considered.

EC4: Completeness: Partially available literature (e.g., abstracts only) is discarded.

3.6. Search Protocol

After the “planning” step focused on defining the search protocol was concluded, we proceeded with the “conducting” step. This was accomplished with no specific tool other than Microsoft Excel spreadsheets, as proposed by Carrera-Rivera et al. [65] for both searching and data extraction. As outlined in [84,85], the removal of irrelevant records was carried out via a two-phase approach to ensure a comprehensive and systematic selection process, where only the most relevant and eligible studies were included in the final review. In the initial stage, titles and abstracts were independently screened by two reviewers to swiftly exclude studies that were clearly irrelevant. The second phase involved a full-text screening, which allowed for a more detailed evaluation and resolution of any ambiguities or borderline cases identified during the abstract screening. Disagreements were addressed via discussion to reach a consensus between the reviewers. The study selection process is illustrated in Figure 1. As discussed, our research design adheres to the PRISMA methodology in accordance with the PRISMA 2020 guideline [69].

3.7. Data Extraction Procedure

This phase focused on creating a data extraction form to record the relevant information from the search strategy [86]. Two reviewers, independently and in duplicate, performed the data extraction procedure from the studies included in the quantitative synthesis (n = 181). The collection form, along with the coding protocol, is included in Appendix A. It comprises the following fields per record: “id”, “title”, “year”, “authors”, “country”, “source”, “publisher”, “category”, “research question”, and “reference”. Disagreements were resolved via discussion and consensus.

All studies were categorized (i.e., “category” field) based on whether their domain focus aligned with one of the following four categories: healthcare market and clinical business research; AI and cloud-native technologies in healthcare; CRIS-CTMS-related literature; and AI applied to medical features in CRIS-CTMS. The justification for the selection of these four domains is grounded in the research motivation presented in Section 3.1. Organizing the relevant studies into these domains enhances the analytical rigor and comprehensiveness of the analysis needed to address the RQs and validate the TPs outlined in Section 3.2.

4. Results and Discussion

This section presents the results and discussion of the SLR. The analysis commences with a detailed account of the processes of literature identification, screening, and selection, including a quantitative characterization of the studies included. The following subsections synthesize key findings related to prevailing market trends, the role of business leadership in AI adoption, and the evolution of AI and cloud-native clinical research information systems. The presentation of results is complemented by a critical assessment of leading AI-powered CRIS-CTMS solution providers and the degree of feature completeness across the sector. The section concludes with an integrative discussion of principal contributions, limitations, and the broader implications of the findings for future research and professional practice.

4.1. Literature Search Results

The search protocol identified 7283 records, which were screened by title and abstract via the search strategy depicted in Figure 1. These included records from both database searches (n = 7182) and handsearching on the Internet (n = 101), as justified in Section 3.4. Based on the inclusion and exclusion criteria, only 4480 records qualified for full-text screening, of which 4299 were excluded due to lack of relevance. A total of 181 records were included in the quantitative synthesis for reference in our study. Figure 2 illustrates the distribution of the SLR results, corresponding to the data sources based on the search strategy.

Table 1. Summary of characteristics of included studies.

Study Characteristic	n (%)
Data sources
Indexed peer-reviewed database	129 (71.27%)
Corporate information	20 (11.05%)
Press releases	21 (11.60%)
Market reports	8 (4.42%)
Article preprints	3 (1.66%)
Publication year
2025	2 (1.10%)
2024	48 (26.52%)
2023	88 (48.62%)
2022	19 (10.50%)
2021	8 (4.42%)
2020	5 (2.76%)
2019	6 (3.31%)
2018	4 (2.21%)
2017	1 (0.55%)
Country
United States	68 (37.57%)
United Kingdom	18 (9.94%)
India	16 (8.84%)
Germany	10 (5.52%)
China	8 (4.42%)
France	8 (4.42%)
Italy	7 (3.87%)
Canada	6 (3.31%)
South Korea	5 (2.76%)
Other countries	23 (19.34%)
Knowledge domain
Healthcare market and clinical business research	31 (17.13%)
AI and cloud-native technologies in healthcare	33 (18.23%)
CRIS-CTMS-related literature	77 (42.54%)
AI applied to medical features in CRIS-CTMS	40 (22.10%)

provides an overview of the characteristics of the studies included in the SLR.

According to the results, more than half of the selected studies (52.5%) were concentrated in the following three major academic databases: ScienceDirect, SpringerLink, and PubMed. Most of the records (75.14%) were published between 2023 and 2024. Additionally, the three countries with the highest number of records (i.e., the United States, the United Kingdom, and India) accounted for 56.35% of the studies included in the qualitative synthesis.

This paper analyzes novel technologies applied to computer-based healthcare information systems and leading AI-powered CRIS-CTMS solution providers. Figure 3 and Figure 4 illustrate how this analysis relates to the studies selected through the SLR.

As shown in Figure 3, most of the examined literature relates to CRIS-CTMS solutions in healthcare companies. However, few studies have examined how these companies are applying AI and cloud-native technologies in their clinical systems. This reflects the overall lack of maturity in digital transformation within the sector, as discussed in Section 4.2. In contrast, Figure 4 highlights that AI-powered CRIS-CTMS suppliers collaborate with pharmaceutical and biotechnology companies, as well as with research institutions. These collaborations involve major organizations such as Bristol Myers Squibb, Johnson & Johnson, AbbVie, Servier Pharma Group, Sanofi, Roche, GlaxoSmithKline, AstraZeneca, Pfizer, Merck KGaA, IMA Group, OncoBay, Johns Hopkins University, and Georgetown University, among others. A detailed review of these initiatives is provided in Section 4.4.

Finally, to further connect the research methodology to the surveyed studies and present the outcome of the SLR, a standardized results spreadsheet was prepared, as shown in

Table 2. SLR results linked to the structure of the paper.

Research Question	Studies	Main Section
RQ1: What are the main growth drivers and challenges in the healthcare industry?	[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,35,36,37,38,39,40]	1. Introduction
RQ2: What opportunities does the adoption of AI and cloud-native computing bring to clinical research, to address the headwinds identified in RQ1?	[41,42,43,44] [29,30,31,32,33,34,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116]	2. Theoretical Foundation 4.2. Market Research and Business Leader Analysis
RQ3: To what degree of maturity are the main pharmaceutical and biotechnology companies adopting those technologies in their CRIS-CTMS solutions?	[117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143]	4.2. Market Research and Business Leader Analysis
RQ4: What are the key feature domains and medical informatics capabilities that AI-powered CRIS-CTMS applications should implement to deliver holistic, advanced clinical research analytics?	[47,48,49,50] [144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176,177]	2. Theoretical Foundation 4.3. AI and Cloud-native Clinical Research Information Systems
	[178,179,180,181,182,183,184,185,186]	5. Conclusions
RQ5: What portfolios, open innovation initiatives, and degree of completeness do leading AI-powered CRIS-CTMS providers have regarding the key feature domains identified in RQ4?	[187,188,189,190,191,192,193,194,195,196,197,198,199,200,201,202,203,204,205,206,207,208,209,210,211,212,213,214,215,216,217,218,219]	4.4. Literature Review of Leading AI-powered CRIS-CTMS Solutions

. This table links the research questions to the related literature and cross-references the sections of the study where each research question is addressed. This facilitates the correlation of the findings with the structure of the paper, supporting traceability.

4.2. Market Research and Business Leader Analysis

The rapid evolution of digital technologies is reshaping the pharmaceutical industry, particularly in clinical research, where AI and cloud-native computing offer significant transformative potential. As outlined in TP3, these technologies aim to address key challenges in clinical trials, such as optimizing patient recruitment and ensuring regulatory compliance. Furthermore, in alignment with TP4, AI and cloud-native solutions play a crucial role in accelerating drug development, reducing costs, and improving treatment accuracy. This section examines how MNEs leverage these advancements to enhance the readiness and maturity of their clinical research information systems.

To achieve this, we examine two main aspects. On the one hand, we survey the overall pharmaceutical market and business outlook on AI investments to address the sector’s challenges. On the other hand, and to address these challenges, we explore the specific actions taken by leading MNE pharmaceutical companies to enhance the readiness and maturity of their computer-based clinical research information systems [143] by utilizing Industry 4.0 technologies [220,221]. Our focus will be on the adoption of AI [87] and cloud-native computing [88], as these technologies have garnered considerable attention in both academia and industry for their application in computer and information systems engineering [89,90,91,92,93].

Since the COVID-19 pandemic, healthcare companies have significantly increased their digitalization investments to increase their operational effectiveness and sustain their productivity [29,30,31]. In 2022, healthcare and pharmaceutical products led AI adoption (15.7%), followed by fintech (13.68% increase), automotive (13.65%), and business and legal (13.6%) [94]. The global market share of AI in healthcare is expected to reach USD 17.8 billion, with a compound annual growth rate (CAGR) of 43.8% from 2019 to 2025. The global outlook for AI in the pharmaceutical market from 2022 to 2030 is illustrated in Figure 5. According to the macro analysis, in the United States, the AI market in the healthcare and pharmaceutical industry exceeded USD 1.15 billion in 2020, with a forecasted CAGR of 44.2% from 2022 to 2027. Additionally, in Europe the AI healthcare and pharmaceutical market surpassed USD 700 million, with a projected growth rate of 43.9% from 2022 to 2027 [95].

This solid addressable market is supported by AI applications in healthcare [96]. These include drug discovery and design for predicting molecule activity, effectiveness, and toxicity on the basis of labeled training datasets [97,98]; quality control in manufacturing using supervised machine learning (ML) models [99] to anticipate production failures and quality deviations [100,101]; patient outcome prediction applying patient labeled datasets and clinical data to forecast disease progression [102,103]; pharmacovigilance to anticipate adverse events in drug administration, which is based on labeled data from patient reports [104,105]; and predictive clinical trial outcomes, using historical labeled data from similar clinical trials and patient characteristics, which can be applied to enhance treatment [106,107]. Other applications include clinical and trial patient recruitment enhancements, via machine learning techniques, which allow the grouping of data points by natural similarity and distance [108,109]; gene expression profiling by means of principal component analysis (PCA) [110,111]; and drug anomaly detection through outlier data points [112,113].

These growing opportunities encourage healthcare companies to increase their AI readiness to avoid falling behind competition [114,115]. On the basis of the research by CB Insights [116] on 50 pharmaceutical companies, Figure 6 illustrates the AI readiness journey of the top 15 organizations, highlighting varying levels of maturity. The AI index score, which is based on CB Insights data, combines the following three factors: talent, execution, and innovation. First, the talent dimension assesses the company’s expertise by evaluating the number of specialized employees and strategic AI-related hires. Second, execution is measured by the organization’s capacity to deliver AI-based products and services to the market, considering business partnerships, product announcements, and earnings reports. Third, innovation is assessed through the analysis of published patents, acquisitions and R&D activities.

Several key findings can be derived from this study. In 2023, only 9 out of the 50 evaluated pharmaceutical MNEs achieved an AI maturity score of 50% or higher. The top three organizations (Roche, Bayer, and Johnson & Johnson) scored below 80% in AI readiness and maturity. Some pharmaceutical leaders sometimes build AI competencies by acquiring digital-born startups. A review of previous research indicates that significant investments and strategic focus will be required by pharmaceutical and biotechnology MNEs over the next five years to meet expectations of exponential growth. In addition, a statistical analysis of the relationships between companies and their AI index scores is presented in

Table 3. Statistical analysis of the 50 pharmaceutical companies examined and their AI maturity levels.

Measure	Value	Analysis
Mean (µ)	29.56	The average AI maturity score aligns with companies like Vertex (22.47) and Sarepta (22.30), indicating that most firms are still in early AI adoption phases.
Median (2nd Quartile)	22.12	Half of the companies, including Alnylam (20.15) and Alkermes (20.07), score below this, showing a clear divide between AI leaders and laggards.
Standard Deviation (σ)	20.94	The high variability suggests leaders (Roche, Bayer) invest heavily in AI, while lower-tier companies like Intra-Cellular Therapies (4.31) and Grifols (5.60) struggle with adoption.
Range	73.17	Roche (77.48) leads AI adoption, while Intra-Cellular Therapies (4.31) ranks the lowest, highlighting extensive disparities in AI maturity across the industry.
Interquartile Range (IQR)	39.09	The middle 50% of companies range from Q1 (9.76) to Q3 (48.85), with firms like Merck (49.13) and Biogen (47.38) significantly ahead of lower-tier firms.
1st Quartile (≤25%)	9.76	Companies like Karuna Therapeutics (9.74) and Legend Biotech (9.72) fall in this range, indicating low AI adoption. Many smaller firms are trailing in AI adoption.
3rd Quartile (≤75%)	48.85	Companies above this, like Pfizer (52.10) and GSK (51.79), are leading in AI-powered drug development and data analytics.

.

To provide context for these findings and a more detailed review of leaders’ maturity levels and actions taken,

Table 4. Innovation activities of selected pharmaceutical and biotechnology MNEs supported by AI and cloud-native technology.

Pharmaceutical/Biotechnology MNE	Innovation Supported by AI and Cloud-Native Technology
Roche (Basel, Switzerland)	Apollo, Roche’s personalized healthcare platform, is built on top of Amazon Web Services (AWS). It delivers capabilities for automated data processing, analytics, and collaboration tools, supporting innovations in cancer diagnosis and early lung cancer detection through AI [117]. In collaboration with AWS and Ibex Medical Analytics, Roche enabled access to Ibex’s AI-driven decision support technology for breast and prostate cancer diagnosis. The system includes AI-supported clinical image analysis [140]. Roche also partnered with Medial EarlySign Ltd., which uses AI capabilities to provide evidence-based machine learning models for lung cancer. These tools support early detection and treatment, potentially improving survival rates [119].
Johnson & Johnson (New Jersey, NY, USA)	Janssen, the pharmaceutical division of Johnson & Johnson, is integrating AI throughout the entire drug development process. This covers AI-driven protein structure prediction, clinical trial design, patient recruitment and manufacturing. It uses digital twin technology [120] for bispecific monoclonal antibody production simulations, and the Trials360.ai platform for end-to-end management of clinical trials [121]. Johnson & Johnson has partnered with Anumana and Mayo Clinic to improve pulmonary hypertension diagnostics. Additionally, the company collaborated with Ultromics Ltd. (Oxford, UK) and Atman Health (Needham, MA, USA) to develop AI models to detect subtleties in echocardiograms for cardiac amyloidosis treatment, where algorithms recommend flagging patients for confirmatory testing [122].
Pfizer (New York, NY, USA)	Pfizer, in collaboration with AWS, has developed a scientific data cloud using generative AI (GenAI) [123] to enhance drug design and discovery. This approach was instrumental in accelerating COVID-19 vaccine approval in 269 days, a process that typically takes between 8 and 10 years [124]. Moreover, in the area of advanced precision medicine and drug discovery, Pfizer is collaborating with Google Cloud by using the following two tools: the Target and Lead Identification Suite, is designed for supporting the prediction and explicability of protein structures, and the Multiomics Suite, which is designed to extract, transform and load genomics data [125]. Additionally, Google Cloud offers AlphaFold2, an ML model for proteomics that runs on Vertex AI, which is Google’s machine learning operations (MLOps) platform [126].
Moderna (Cambridge, MA, USA)	Moderna Inc. is a biotechnology company focused on messenger RNA (mRNA)-based vaccines. In the fight against COVID-19, Moderna, like Pfizer, leveraged AWS cloud resources and AI capabilities to model the virus in just two days [121]. The partnership with AWS has enabled successful projects related to real-world evidence data extraction and analysis, AI-powered customer experience improvements, and drug discovery and commercialization based on Amazon Data Exchange and Amazon Redshift [127]. In the area of clinical research, Moderna partnered with Google Cloud for easier and faster access to actionable insights, supporting real-time decisions and improving trial visualization by using the Google Looker tool [128].
AbbVie (North Chicago, IL, USA)	Together with the protein therapeutics company BigHat Biosciences (San Mateo, CA, USA), a partnership for the discovery and development of antibodies in oncology and neuroscience using AI has been established [129]. AbbVie has implemented AI-powered tools like Cortellis Search for preparing clinical report studies (CSR) in minutes. Another example is FOCAL, a framework for clinical AI language supporting content identification and translation that assists in clinical trial execution [130]. Additionally, AbbVie collaborates with Intel on AbbVieFish Machine Translation, based on the Transformer natural language processing (NLP) model [131]. This is used to scan research documents enabling the discovery of new treatments and manufacturing methods [132].
Gilead Sciences (Foster City, CA, USA)	As part of an open innovation initiative, Gilead Sciences partnered PathAI (Boston, MA, USA), an AI-powered company specializing in pathology research, to develop an AI-based tool for scoring nonalcoholic steatohepatitis (NASH). This tool correlates with pathologists’ scores and biomarkers for patient outcome forecasting [142]. For data management, the biotechnology company uses an Intelligent Data Management Cloud from Informatica, an enterprise cloud data management provider. This approach implements a data mesh framework for self-service access to vast amounts of data, which is used in drug design, discovery, and commercialization [133]. Additionally, Gilead Sciences collaborates with AWS, adopting Amazon Redshift as a cloud-based data warehouse solution for third-party medical use cases. The proposed architecture allows serverless data ingestion, consumption and built-in analytics [134].
Regeneron (Tarrytown, NY, USA)	Databricks Inc., a leading supplier of solutions for massive parallel processing, provides Regeneron with a data science platform running on AWS for enhanced productivity. It uses Apache Spark [135] powered pipelines for drug target identification and minute-scale extraction, transformation and loading operations [136]. Regeneron has multiple ML models used in drug design and discovery, including image detection of protein droplets and prediction of immune responses [137]. Additionally, one of the pharmaceutical company’s most ambitious initiatives is the creation of the Regeneron Genetic Center, one of the largest genomic databases in the world, which, by using proprietary data analytics and machine learning, effectively analyzes relationships and associations among genes and diseases [138].

outlines noteworthy innovation activities related to AI and cloud technology adoption by pharmaceutical and biotechnology MNEs.

4.3. AI and Cloud-Native Clinical Research Information Systems

The previous state-of-the-art review on AI maturity in pharmaceutical and biotechnology companies highlights their ongoing transformational journey. These companies increasingly view themselves as clinical data organizations, reflecting the broader industry shift toward data-driven decision-making. However, AI and cloud-native computing adoption remain in early stages across the industry, as anticipated in TP5. Additionally, in line with TP6, adoption rates correlate with organizational scale, with larger companies advancing more rapidly. These findings underscore the need to develop strategic frameworks to facilitate broader technological transformation within the sector.

A cornerstone of this transformational journey is the adoption of advanced clinical research information systems (CRISs) that integrate AI and cloud-native technologies [177]. AI technologies enable efficient management of massive clinical datasets, support forecasting, decision-making, and data inference [144,145]. Cloud-native applications, on the other hand, are capable of delivering global scalability through elasticity and high-availability patterns [146]. Together, they can remove information silos while increasing productivity, reducing pharmaceutical costs, and enhancing healthcare quality [147,148]. A CRIS is a broad term used to define a computer-based software appliance managing clinical research tasks, encompassing medical informatics capabilities for business processes, information flows, patient records, and data analytics [143,149]. Examples include electronic data capture (EDC) [150], electronic health records (EHRs) [151], clinical research data warehouses (CRDWs) [152], and clinical trial management systems (CTMSs) [153].

When applied to a CRIS, AI and cloud-native technologies serve as powerful tools, providing transformative opportunities and significant enhancements across clinical application domains. These include, but are not limited to, drug discovery and design; quality control in drug manufacturing; patient outcome prediction; pharmacovigilance and anticipation of adverse events; clinical trial outcome forecasting; enhanced patient recruitment; gene expression and biomarker profiling; and predictive drug anomaly detection. A CRIS acts as the operational backbone of the AI ecosystem for a clinical data organization. All the actions listed in Table 4 pivot around a CRIS. These activities provide data that support information sharing between investigators, clinical event tracking, patient follow-up, advanced analytics, modeling, and optimization of clinical pathways [154].

As previously noted, a CTMS is a specific implementation of a CRIS tailored to clinical trials [153,154]. The industry and research interest behind CTMS engineering is vibrant, particularly when leveraging AI capabilities. According to Sudhakar et al. [155] and Precedence Research [139], the market for AI-based clinical trials is projected to grow from USD 1.6 billion in 2023 to USD 18.9 billion by 2033, with a 20% CAGR during 2024–2033, driven by North America, Europe, and the Northwestern Pacific. In this context, AI-powered CTMS solutions are key to supporting the successful delivery of this projected market growth. They are essential for delivering efficient, auditable, and high-quality clinical trial outcomes while optimizing the overall process management of clinical trials [156,157]. Given their significance in both business and information system research, the remainder of this manuscript focuses on CRIS engineering applied to clinical trial management systems (CRIS-CTMSs).

As with any health system, CRIS solutions must be devised holistically, considering the interconnections between their capabilities to address complex healthcare structures [158]. Surveys by Sampson et al. [159] and Zhuang et al. [160] reveal that the CRIS-CTMS marketplace is fragmented due to the lack of comprehensive feature domains. This results in potential data inconsistency across multiple vendor solutions. To address this, Figure 7 introduces a novel 360-degree blueprint defining essential medical feature domains for future AI-powered CRIS-CTMS implementations. This holistic approach aims to drive the design and delivery of advanced data analytics for clinical trials, avoiding scattered solutions and information silos. Our contribution also lays the foundation for the maturity assessment of CRIS-CTMS solutions.

The primary motivation behind these feature domains is to enable the thorough oversight of clinical research in a cohesive manner, enhancing trial performance through improved quality and operational efficiency [161,162]. These domains, which are essential for comprehensive trial supervision, require advanced CRIS-CTMS solutions. The definitions of the proposed feature domains are derived from the relevant literature, ensuring smooth system integration, as follows: According to McCaskell et al. [51], the study initiation and trial management and monitoring functions are essential components. Singh et al. [52] emphasized the importance of trial design as a preliminary stage, which facilitates the streamlining and cost-effectiveness of trial execution. Kandi and Vadakedath [53] highlighted the necessity of clinical trial data analysis and regulatory compliance audits. Data reporting, as discussed by Rombach et al. [54], is crucial for producing valid and reliable conclusions from clinical trials. Finally, based on the findings of Copland et al. [55], supply management has also been identified as a critical domain feature to ensure timely and efficient delivery of investigational products necessary for study execution. These findings align with the definitions outlined in TP7 and TP8, highlighting a gap in the literature regarding cohesive and seamless integration of medical features across all trial domains. The following subsections describe the feature domains that novel AI and cloud-native CRIS-CTMS solutions should address.

A.

Clinical Study Design

It encompasses the study design for both observational and interventional studies. The clinical research protocol is designed to evaluate new drugs, therapies, or medical devices, including patient outcomes, quality of life, and impacts on medical practice. Precise translation of the clinical protocol is essential for trial success. This prevents protocol deviations, including key aspects such as data and privacy regulations. Harrer et al. [20] and Tsuchiwata and Tsuji [163] focused their research on applying AI to reduce the need for sample collection and relax strict patient recruitment ratios without affecting the accuracy of drug development. However, the scope of these previous works is limited to clinical study design and does not extend to other feature domains.

B.

Clinical Study Initiation and Setup

This feature domain focuses on the preparation and readiness for the clinical data management plan (CDMP) that precedes trial management and execution. The CDMP is composed of an electronic case report form (eCRF) for collecting patient data; eCRF collection guidelines; a standard data tabulation model (SDTM) mapping for representing trial data in a standardized manner; and trial database creation and setup, which will support data queries from the investigators. In the same vein, Matsuzaki et al. [164] designed a model for integrating pre-existing databases through an SDTM. The proposed use of AI focuses on suggesting metadata based on the therapeutic area and is limited to this feature domain.

C.

Trial Management

It covers all the capabilities required for the operational execution of clinical trials, from initiation to closure. This feature domain encompasses planning and supervising clinical sites, patient recruitment, continuous oversight through monitoring visits and report compilation, protocol deviation management, and clinical project issue and risk management. Several studies applying AI to trial management have been recently published. Alexander et al. [165] presented an AI-powered design for patient matching as part of patient recruitment in lung cancer trials. Ferdowsi et al. [166] investigated the application of AI technologies limited to trial protocol deviation predictions based on risk-related metrics.

D.

Supply Management

Through highly accurate forecasting techniques applied to drug supply for patients, medicine prescriptions during clinical trial treatments can be optimized, avoiding the risk of inventory shortages and reducing supply costs. Related work has focused primarily on applying ML regression and forecasting techniques for supply management to anticipate medical demand. Liu [167] presented a model for medical consumables via a dynamic neural network for application during the COVID-19 pandemic, demonstrating superior results compared to intermittent demand methods. Rekabi et al. [168] devised a pharmaceutical supply chain approach using AI-based linear and quadratic models to predict drug demand.

E.

Data Management

This feature domain covers the collection, cleaning, validation, and visualization of data gathered from the clinical trial without omissions. Its objective is to generate high-quality and reliable outcome data in compliance with the protocol design and regulatory standards. This includes smart, advanced queries on trial data, including query management and validation, term categorization for analysis via medical coding, data consistency, and adverse event reporting. Related work includes the proposal by Yang and Kar [105] for the early detection of adverse drug reactions using supervised ML models and the contribution by Edin et al. [169] for a new ML-based model of medical coding and parameter optimization.

F.

Data Analysis and Reporting

During the course of a clinical trial, extensive amounts of data are created, all of which require structured interpretation. AI techniques effectively generate actionable insights, consolidate data, and boost the review process via dashboards and advanced visualization capabilities. They also provide automated report generation for interim analysis, data transformation and harmonization, collaborative review, and pharmacovigilance. Related works in this feature domain include that of Gómez et al. [170], who proposed an analytics framework for visually exploring longitudinal data. Additionally, Jing et al. [171] designed a visual analytics interface for filtering and summarizing large clinical datasets.

G.

Regulatory Filings and Submission

Regulatory work includes preparing submissions to regulatory authorities and ethics committees in compliance with local regulations. The required documentation includes non-disclosure agreements, confidential disclosure agreements, contract trial agreements, replies to queries made by regulatory bodies, and post-approval submissions. The electronic trial master file (eTMF) [172] is built and delivered to sponsors. AI helps streamline these processes by reducing time, improving accuracy, and increasing regulatory compliance. Patil et al. [173] and Derraz et al. [174] identified current regulatory limitations in implementing AI-based filing systems and analyzed the challenges involved in bridging this gap.

Finally, the proposed 360-degree reference blueprint has been connected to good clinical practice (GCP), as shown in

Table 5. Mapping between 360-degree reference blueprint and GCP principles.

Feature Domain	Relevant GCP Principles [175]	Rationale
Study design	- Compliance with protocol (GCP 2.5) - Trial design (GCP B.4)(GCP 2.5) - Quality management (GCP 3.10)	Ensures the study is scientifically sound, adheres to the approved protocol, follows a structured design, and maintains high-quality standards throughout its execution.
Study initiation and setup	- Investigator qualifications and training (GCP 2.1) - Investigator selection (GCP 3.7) - Risk management (GCP 3.10.1)	Guarantees that qualified investigators conduct the trial, appropriate sites are selected for effective execution, and potential risks are proactively identified and mitigated.
Trial management	- Data governance (GCP 4) - Monitoring (GCP 3.11.4) - Safety reporting (GCP 2.7.2)	Maintains secure and accurate management of trial data, facilitates continuous oversight to identify and resolve issues promptly, and upholds participant safety through timely adverse event reporting.
Supply management	- Manufacturing, Packaging, Labeling, and Coding Investigational Products (GCP 3.15.2) - Supplying and handling investigational product (GCP 3.15.3) - Investigational product accountability at investigator site (GCP C3.1)	Supports proper manufacturing, labeling, distribution, and tracking of investigational products in compliance with regulatory requirements, thereby maintaining product integrity and trial reliability.
Data management	- Validation (GCP 4.3.4) - Data handling and record-keeping (GCP B.14) - Relevant metadata, including audit trails (GCP 4.2.2)	Strengthens the reliability and security of trial data through system validation, proper documentation, and traceable audit trails that support data integrity and regulatory compliance.
Data analysis and reporting	- Data handling (GCP 3.16.1) - Statistical programming and data analysis (GCP 3.16.2) - Data report (GCP 3.17)	Assures accurate data processing, the application of appropriate statistical methodologies, and the transparent reporting of findings in accordance with scientific and regulatory standards.
Regulatory filings and submissions	- Submissions and communications (GCP 1.1) - Notification/Submission to Regulatory Authority (GCP 3.8.1) - Auditing processes (GCP 3.11.2.2)	Facilitates timely and accurate regulatory submissions, effective communication with authorities, and adherence to audit requirements, supporting compliance and accountability.

. The GCP defines statements and recommendations to optimize patient care, providing a unified standard to facilitate acceptance of clinical research and trials [175]. The intention is twofold. On the one hand, validating the framework against GCP ensures clinical trials adhere to regulatory obligations and ethical practices [176]. By making such a connection, the adherence of our proposal to good clinical practice can be assessed in a transparent manner. On the other hand, it facilitates adoption and application of the theoretical contribution by the community. As clinical practitioners are used to the GCP, the proposed blueprint will allow them to easily map their work to the medical features in the CRIS-CTMS.

4.4. Literature Review of Leading AI-Powered CRIS-CTMS Solutions

Once the reader has been presented with the proposed blueprint for advanced CRIS-CTMS, this section examines the leading industry suppliers of AI-powered CRIS-CTMS solutions. This review includes their collaboration initiatives with pharmaceutical and biotechnology leaders aimed at shaping the future of clinical research. These suppliers offer solution portfolios featuring advanced medical capabilities powered by AI and cloud-native technologies, enabling breakthrough solutions validated in collaboration with pharmaceutical companies, as highlighted in TP9. However, their medical capabilities are often fragmented, focusing on one or a few medical domains only, as noted in TP10. This indicates a clear need for further development to create more comprehensive, integrated solutions.

In recent years, several key companies have successfully applied AI to clinical trial research by offering their products as software as a service (SaaS) on hyperscale cloud providers, such as AWS, Google Cloud, or Microsoft Azure. This cloud model reduces upfront capital expenditure (CAPEX) by transitioning costs to operational expenditure (OPEX) on a pay-as-you-go basis, enabling faster market introduction and scalable growth alongside demand. To illustrate the industry momentum, and based on market research by Elflein [187] in Figure 8, we highlight the AI-powered CRIS-CTMS solutions that raised the largest venture capital funding in 2023.

Saama Tech leads the AI and cloud-native clinical trials market. Based in Silicon Valley, the firm provides an AI-driven platform with more than 90 ML models trained on over 300 million data points [188]. Its portfolio includes clinical operations and finance modules, operational and patient insights, SDTM submission management, and smart queries, reportedly reducing processing time by 90%. Saama Tech raised USD 0.5 billion in 2023, the largest amount in this ranking.

The second group includes Owkin, ConcertAI, and PathAI, which collectively raised USD 0.86 billion in 2023. Owkin is an AI-based biotechnology company founded in 2016 in Paris, France [191]. It delivers AI-powered CRIS-CTMS capabilities in a SaaS portfolio, focused on drug development, diagnostic biomarker prescreening, outcome prediction, and drug discovery. In 2023, the European Medicines Agency (EMA) approved the use of Owkin’s deep learning models for covariate adjustment in oncology trials [192]. In contrast, ConcertAI leverages more than 150 ML models and is a leading provider of AI-powered clinical services centered on their CRIS-CTMS solution, focusing primarily on oncology research in observational studies [193]. This type of study is key in clinical research, allowing healthcare professionals to make informed decisions by using real-world evidence to validate the safety and efficacy of medical products. Additionally, PathAI, a vendor founded in 2016, offers an AI-powered CRIS-CTMS portfolio specialized in pathology investigations. According to the supplier, approximately 90% of pharmaceutical and biotechnology companies employ their solution for pathology-related clinical research [194]. The vendor offers more than 20 ML algorithms focused on end-to-end AI digital pathology workflows and graph neural networks (GNNs) [195] alongside GenAI. The commercial offer includes SaaS products and services for oncology and AI-based measurement of nonalcoholic steatohepatitis (AIM-NASH) and inflammatory bowel disease.

The third group of AI-powered CRIS-CTMS solutions includes Lantern Pharma, Unlearn, and AiCure, which together raised USD 0.23 billion in 2023. Lantern Pharma [196] demonstrates an example of backward vertical integration as follows: as a pharmaceutical company, Lantern Pharma developed an AI-powered CRIS-CTMS, called the Response Algorithm for Drug Positioning and Rescue (RADR), which includes over 60 billion oncology data points and more than 200 ML algorithms for personalized cancer treatments. Unlearn uses AI to minimize trial errors and observe treatment effects in early-stage studies without adding more patients [197]. Its AI-powered CRIS-CTMS portfolio focuses on a digital twin model [86] for clinical trials in neuroscience, immunology, cardiology, and metabolic and musculoskeletal areas. Unlearn also provides professional services to support regulatory agency interactions and statistical analysis. Finally, AiCure [198], founded in 2010 and based in New York, offers an AI-powered CRIS-CTMS portfolio aimed at enhancing drug development by ensuring that clinical trials capture meaningful data. AiCure’s solutions use a patient-centric approach, supervising medication adherence through patient facial recognition and monitoring treatment based on the analysis of behavioral biomarkers.

Table 6. Survey of leading AI-powered CRIS-CTMS solutions (Part 1 of 2): portfolio and open innovation collaborations.

AI-Powered CRIS-CTMS Solution	Portfolio	Open Innovation Collaborations
Saama Tech (Campbell, CA, USA)	Data Hub centralizes clinical, operational and financial data by ingesting, centralizing and standardizing data sources. Operational Insights provides holistic clinical operations views. Patient Insights enables the visualization and prediction of patient data. Source of Submission tool simplifies sponsor submissions. Smart Data Quality accelerates data review, identifying data discrepancies with real-time data cleaning [189]. For the portfolio realization, Saama Tech applies deep learning, predictive AI, and ML-based analytics to enhance drug efficiency and ensure patient safety, in addition to utilizing NLP to expedite the extraction of insights from clinical trial documents and data analysis [190,199].	In October 2023, AstraZeneca announced a multiyear contract with Saama for accelerating drug development by modernizing data and medical review processes [190]. In 2020, Pfizer partnered with Saama to speed up the COVID-19 vaccine by using Smart Data Quality to clean data from over 30,000 patients [188]. This contract was renewed in 2024 to expedite regulatory submissions using Saama’s advanced biometrics research and analysis network.
Owkin (Paris, France)	Using federated learning [144] via Substra [200], Owkin offers advanced solutions for drug discovery and development. These solutions use ML preclinical models, AI-based external control arms, enhanced inclusion criteria models, and prognostic biomarker covariance adjustment. This includes ML algorithms for biomarker analysis, including prescreening, identification, and outcome prediction. Owkin’s portfolio includes deep learning models, such as convolutional neural networks (CNNs), to analyze digitized hematoxylin, eosin, and saffron (HES)-stained whole slide images (WSIs) for predicting patient outcomes and detecting genetic mutations [191].	In 2021, Owkin and Sanofi entered a 3-year contract to develop oncology-focused predictive ML models [201]. Owkin received an USD 80 million payment from Bristol Myers Squibb in 2022, as part of a USD 480 million deal for improving efficiency and precision for clinical trials [191]. France’s Servier Pharma Group closed a contract in 2023 with Owkin to identify tumor characteristics and patient subgroups [202].
ConcertAI (Cambridge, MA, USA)	ConcertAI’s AI-powered CRIS-CTMS portfolio in real-world evidence features two main offerings: Digital Trial Solution and Clinical Trial Optimization 2.0. Digital Trials Solution enhances trial feasibility, site selection, and documentation management with real-time reporting. Clinical Trial Optimization 2.0 improves study inclusion/exclusion criteria, site identification, building trial strategies, and assessing study feasibility [193]. Additionally, ConcertAI has expanded its portfolio with CARA AI, a digital platform providing both predictive and generative AI capabilities to support clinical site selection. By leveraging multimodal and predictive AI inference, this solution simulates potential challenges researchers may encounter during the selection, thereby helping expedite decision-making [215].	Bristol Myers Squibb partnered with ConcertAI to improve real-world evidence and observational studies by enhancing data quality and clinical workflows [202]. In 2021, ConcertAI partnered with Janssen to improve diversity in clinical trials to better understand treatment effects in population groups [204]. More recently, ConcertAI and AbbVie signed a multi-year contract where the provider offers its large-scale oncology database to develop cancer drugs [219].
PathAI (Boston, MA, USA)	The AI-powered CRIS-CTMS platform by PathAI, named AISight, is a cloud-based SaaS solution for pathology operations. It supports data collection, response monitoring, quality control, and analysis. It features AI-based measurement of nonalcoholic steatohepatitis (AIM-NASH), tumor cellularity quantification, and molecular prediction. The ML-based algorithm uses supervised learning for patient identification with molecular status prediction, biomarker quantification and endpoint assessment [194]. In addition, PathAI’s portfolio incorporates advanced deep learning technologies for computer vision, utilizing CNNs. This enables precise and efficient assessment of tissue samples for diagnostic and prognostic purposes in digital pathology [216].	In 2022, PathAI and Bristol Myers Squibb contracted for translational research in cancer research, immunology, and fibrosis [205]. PathAI and GlaxoSmithKline partnered to improve cancer research with AI, focused on liver biopsy, as well as AIM-NASH [206]. In February 2024, PathAI announced an exclusive collaboration with Roche to develop AI-enabled digital pathology algorithms for companion diagnostics [118].
Lantern Pharma (Dallas, TX, USA)	The RADR platform, Lantern Pharma’s AI-powered CRIS-CTMS, enables predicting patient response to new drugs and supports drug development in oncology. RADR uses transcriptome and genomic data, accessing datasets applied to supervised ML models. Primary uses include genomic feature selection and biomarker extraction, preclinical hypothesis generation and validation, patient stratification, and clinical study design. RADR applies predictive modeling for forecasting drug response and patient outcomes, as well as supervised and unsupervised learning for biomarker discovery, patient stratification, and identification of drug-tumor interactions [196].	Lantern Pharma collaborated with Georgetown University for pancreatic cancer research [207]. Johns Hopkins Pediatric Oncology Division and Lantern Pharma partnered for malignant gliomas research using RADR [208]. TTC Oncology and Lantern Pharma signed an agreement in 2023 to research breast cancer, with RADR used for biomarker identification, characterization of drug mechanisms, and treatment discovery [209].
Unlearn (San Francisco, CA, USA)	Unlearn’s TwinRCT platform creates patients’ digital twins to simulate and forecast health outcomes. The supplier has also developed PROCOVA, a statistical methodology for incorporating prognosis scores from digital twins into the trial design and analysis. Unlearn provides disease-specific digital twin models for several therapeutic areas, such as neuroscience and immunology, cardiovascular, metabolic, and musculoskeletal medicine [197]. Based on its portfolio, Unlearn delivers AI-generated digital twins that have demonstrated the ability to improve the efficiency of randomized clinical trials [217].	Unlearn and Merck KGaA announced a multiyear contract in 2022 to accelerate drug development. The focus of the partnership was on immunology trials, where the use of TwinRCT was expected to reduce control arm sizes by 30% or more [210]. In 2023, Unlearn partnered with QurAlis Corporation (Cambridge, MA, USA), a clinical-stage biotechnology company. This collaboration aimed to predict amyotrophic lateral sclerosis outcomes using TwinRCT to produce regulatory-suitable evidence [211].
AiCure (New York, NY, USA)	AiCure has built an AI-powered CRIS-CTMS with a patient-centric approach. The AiCure Site Dashboard provides site coordinators with insights related to patient behavior. The AiCure Platform enables digital biomarkers analysis. AiCure Clinical Site Services supports patient engagement, data-driven guidance, monitoring of metrics, compliance, and data management. Patient Connect offers real-time dosing instructions and intake supervision via facial recognition [198]. AiCure’s portfolio utilizes predictive ML models to forecast adherence risks, enabling more effective patient support and intervention. Additionally, computer vision and multimodal AI technologies are employed to improve medication adherence by analyzing video and audio data to monitor patient adherence and engagement in clinical trials [218].	In 2019, Syneos Health partnered with AiCure via a multiyear contract to incorporate its technology in clinical trials. aiming to capture, predict, and influence patient behavior [212]. AiCure also collaborates with OncoBay to enhance patient quality of life through remote treatment monitoring via smartphones [213]. Together with the IMA Group, and through their Patient Connect product, AiCure provides the ability to monitor patient adherence across different therapeutic areas [214].

presents a review of the product and service portfolios of these solutions, along with the analysis of recent open innovation initiatives in collaboration with pharmaceutical and biotechnology companies. The completeness of each AI-powered CRIS-CTMS solution, regarding the medical feature domains for each supplier implementation, is presented in

Table 7. Survey of leading AI-powered CRIS-CTMS solutions (Part 2 of 2): medical feature domains and maturity scope, calculated as the number of feature domains and degree of feature fulfillment (partial: 0.5, high: 1) according to the 360-degree blueprint in Figure 7.

AI-Powered CRIS-CTMS Solution	Feature Domains	Maturity Score: n (%)
Saama Tech	Study Initiation and Setup: SDTM transformation Supply Management: financial management Trial Management: operational trial execution, streamlined medical review process Data Management: query management, data source verification, data cleaning, predict participant behavior Data Analytics and Reporting: data analytics, patient data visualization Regulatory Filings and Submissions: generation of submission-ready tables, listings, and figures	6/7 (85.71%) All medical features are covered in the fulfilled domains
Owkin	Study Design: optimize endpoint definition, inclusion criteria models Data Management: early estimates of the treatment effect Data Analysis and Reporting: covariance adjustment	2.5/7 (35.7%) Most medical features are covered in the fulfilled domains, with limited coverage in Data Analysis and Reporting
ConcertAI	Study Design: study inclusion/exclusion criteria Study Initiation and Setup: feasibility and site selection, site identification Regulatory Filing and Submissions: clinical trial documentation management Trial Management: real-time reporting, building trial strategies, assessing study feasibility to help predict study performance at sites	3.5/7 (50.7%) Most medical features are mostly covered in the fulfilled domains, with limited coverage in Study Design
PathAI	Study Design: tumor cellularity quantification, molecular prediction algorithms, biomarker quantification, and exploratory endpoint assessment Data Analysis and Reporting: data ingestion, response monitoring, quality control, and exploratory analysis	2/7 (28.57%) All medical features are covered in the fulfilled domains
Lantern Pharma	Study Design: patient stratification Data Analysis and Reporting: potential response prediction to new drugs	1/7 (14.28%) Limited feature coverage in Study Design, and Data Analysis and Reporting
Unlearn	Study Design: patients’ AI-generated digital twins to simulate and forecast health outcomes; disease-specific models for generating patients’ digital twins Data Analysis and Reporting: statistical methodology for incorporating prognosis scores from trial patients’ digital twins into the clinical trial design and analysis	1.5/7 (21.43%) Limited feature coverage in Data Analysis and Reporting
AiCure	Trial Management: patient engagement; real-time dosing for patients Data Analysis and Reporting: data analytics; patient data visualization; dashboard provides insights related to patient behavior; metrics monitoring (i.e., adherence, compliance)	2/7 (28.57%) All medical features are covered in the fulfilled domains

and analyzed using the novel 360-degree maturity blueprint in Figure 7. A maturity score is then calculated based on the number of feature domains covered and the degree of feature fulfillment (ranked as partial: 0.5 and high: 1) by each implementation.

As detailed in Table 6, the surveyed CRIS-CTMS platforms employ advanced deep learning architectures, most notably computer vision and CNNs, to process high-dimensional data. This includes digital pathology images and unstructured clinical documentation. Across numerous disciplines, deep learning techniques have outperformed traditional ML approaches, demonstrating superior inference generalization, enhanced accuracy, and more effective extraction of actionable insights [222,223]. These advantages are especially critical for applications involving complex image and sequence data, which are prevalent in medical informatics. Similarly, the adoption of multimodal AI allows vendors such as ConcertAI and AiCure to integrate and concurrently analyze heterogeneous data modalities including clinical, operational, and imaging data. This comprehensive data combination significantly enhances predictive fidelity and robustness, exceeding the performance boundaries typically associated with classical ML approaches. Federated learning, as exemplified by Owkin’s Substra platform, facilitates distributed model training across disparate institutional datasets without necessitating direct data sharing. This paradigm supports stringent data privacy and regulatory compliance requirements while also improving model generalizability. When predictive inference is implemented within multimodal and federated frameworks, it enables precise, patient-specific risk stratification in real-world clinical settings. Additionally, the application of digital twin technology, such as Unlearn’s TwinRCT, supports the development of advanced simulation models for forecasting clinical outcomes and optimizing trial designs.

From a commercial perspective, these CRIS-CTMS solutions are predominantly delivered as cloud-based SaaS platforms via major public cloud providers including AWS, Microsoft Azure, and Google Cloud. This delivery model ensures on-demand scalability, optimized total cost of ownership, and rapid deployment cycles. For instance, Saama Tech provides full integration with both AWS and Microsoft Azure Marketplaces. At the time of writing, the Saama Platform Enterprise Pilot Study is available on AWS Marketplace at a price point of USD 300,000 per month. Additionally, Saama’s Real World Analytics and Patient Experience modules may be deployed on virtual machines in Microsoft Azure via a bring-your-own-license (BYOL) model. Comparable pricing and architectural deployment details for other vendors are not publicly disclosed. More granular technical specifics, such as infrastructure requirements, identity and access management configurations, performance instrumentation practices and results, and update or release schedules, are not made publicly available and therefore fall outside the scope of this review.

4.5. Discussion and Contributions

Research on AI and cloud-native computing in the pharmaceutical and biotechnology industries has progressed significantly in recent years. However, a systematic review of the justification and maturity of AI adoption, connected to strategic roadmaps for advanced CRIS-CTMS transformations, is still lacking. Existing studies have explored AI and cloud-native technologies in specific therapeutic domains but lack a seamless, structured framework for assessing AI adoption maturity in next-generation clinical information systems. Additionally, the literature on CRIS-CTMS functionalities remains fragmented, offering insights into challenges and advancements but failing to provide a comprehensive blueprint for evaluating digitalization in clinical research. Addressing these gaps is essential for promoting integrated AI and cloud-native adoption, thereby advancing clinical information systems toward greater innovation, efficiency, and patient care.

As explained in Section 2, this work builds upon the existing literature in medical informatics by addressing these challenges. This study contributes a rigorous market analysis of challenges and recent developments within the healthcare industry. To this end, it updates the progress and maturity level of AI and cloud-native technologies adopted by pharmaceutical and biotechnology companies transitioning into clinical data organizations. It examines the medical informatics capabilities of major CRIS-CTMS solutions, including their maturity level based on a 360-degree blueprint, and it highlights collaborations with leading pharmaceutical and biotechnology companies. A SLR has been adopted as the research method, including 181 records in quantitative synthesis. It supported insights from both the scientific and business domains, offering a timely examination of the opportunities arising from AI and cloud-native technologies in CRIS-CTMS solutions.

The primary limitation of our SLR is the scarcity of comprehensive studies and the limited availability of research on the application of AI to CRIS-CTMS solutions. To the best of the authors’ knowledge, this gap can be attributed to the early stage of AI maturity within pharmaceutical and biotechnology organizations. Consequently, the current landscape has yet to generate sufficient momentum to stimulate the production of robust research evidence in this domain. However, as AI development and readiness continue to progress within these sectors, it is anticipated that additional research will emerge. This progress will enable further SLRs on this topic in the coming years. Importantly, the anticipated growth in research output, when combined with the integration of comprehensive academic databases such as Scopus and Web of Science into the search strategy, is expected to enhance the quality and breadth of future literature reviews. As the body of literature expands in line with growing AI adoption in the healthcare industry, leveraging these databases will become increasingly crucial. This approach will help ensure that subsequent SLRs accurately reflect the evolving state of knowledge at the intersection of AI and CRIS-CTMS solutions.

Through a comprehensive market and business analysis, our study examines both quantitative and qualitative factors, along with the challenges facing the pharmaceutical and biotechnology industry. Special attention is given to R&D spending, the lack of return on investment, and the headwinds in clinical research. The focus on R&D spending aligns with TP1, emphasizing global healthcare needs and the pressure on pharmaceutical companies to innovate in response to the aging population, chronic diseases, and the increasing demand for personalized medicine. The industry also faces significant challenges, including rising development costs, regulatory complexities, and inefficiencies in clinical trials. These challenges relate to TP2, highlighting the increasing complexity in drug development and clinical research.

AI and cloud-native computing present transformative opportunities to enhance clinical trials and drug development. AI can optimize the clinical trial process by improving patient recruitment, ensuring better alignment with eligibility criteria, and facilitating regulatory compliance in accordance with TP3. Moreover, the ability of AI to reduce trial time and costs and improve treatment accuracy supports TP4, streamlining the drug development lifecycle and enhancing patient outcomes.

The maturity of AI adoption among leading healthcare companies remains low, in alignment with TP5. To explore this, a survey on AI and cloud-native computing in clinical research was conducted, analyzing market potential through 2032. Our contribution includes a maturity review of companies transitioning into clinical data organizations, emphasizing the need to overcome the maturity gap to unlock the potential of AI in clinical trials. We observed that larger companies lead in AI adoption, as anticipated by TP6.

To address the gap identified in TP7, which highlights the need for a holistic view of medical features in clinical information systems, this research provides a 360-degree maturity analysis of AI-powered CRIS-CTMS solutions. It offers a comprehensive framework to evaluate medical features across various clinical trial domains. Our contribution includes a review of the literature on CRISs, justifying the importance of advanced computer-based clinical trial management systems, which we term AI and cloud-native CRIS-CTMS. We present an original 360-degree blueprint of maturity analysis based on medical features, providing both theoretical and practical guidance for future developments. This work aligns with TP8, emphasizing seamless integration of medical features across all domains of clinical trials. This covers the study design for regulatory filing and submissions, including initiation and setup, trial management, and supply management to data management, and data analytics and reporting. This integrated approach provides a comprehensive reference blueprint for evaluating existing AI-powered CRIS-CTMS solutions. This framework not only enables a more cohesive analysis of current suppliers’ offerings but also aids in enhancing their portfolios. Furthermore, the reference blueprint offers insights that can guide future developments in AI and cloud-native technologies. This study aims to improve clinical trial management and support better healthcare outcomes, more effective treatments, and personalized care solutions.

The expanded research on leading AI and cloud-native CRIS-CTMS solutions reveals that while these providers offer advanced capabilities, their offerings remain fragmented, as anticipated in TP9 and TP10. Although AI-powered CRIS-CTMS providers deliver portfolios with advanced medical features, these solutions are not yet fully integrated across all clinical trial domains, limiting their effectiveness. Some providers have made significant strides, but their medical capabilities are often confined to specific therapeutic areas, indicating a need for broader development. In this context, none of the surveyed vendors completely fulfilled the maturity analysis. Saama Tech leads the market with an 85.71% score, offering the most holistic portfolio, with integrated solutions covering multiple aspects of clinical trials. Nevertheless, the supplier did not provide study design capabilities at the time of our study. ConcertAI and Owkin, as technology-following vendors, ranked 50.7% and 35.7%, respectively, revealing gaps in their portfolios that may require complementing their portfolios with other clinical solutions. Despite progress, all surveyed companies face challenges in achieving seamless medical integration across all trial domains, which limits their potential to fully transform clinical research. The study also reviewed the main open innovation activities of these vendors over recent years. We examined the collaborations between these suppliers and top healthcare companies. Saama Tech partnered with AstraZeneca and Pfizer. Similarly, ConcertAI collaborated with Bristol Myers Squibb and AbbVie. Owkin, in turn, collaborated with Sanofi, Bristol Myers Squibb, and Servier Pharma Group.

The study underscores the need for further development in these portfolios to create advanced, integrated solutions that cover the entire lifecycle of clinical trials. By addressing these gaps, our research provides valuable insights into the future of AI and cloud-native technologies in clinical research. The findings provided in this paper offer a timely and strategic roadmap for fostering innovation and enhancing efficiency across the healthcare ecosystem.

5. Conclusions

Health and care systems are fundamentally interconnected in terms of services and capabilities [158,224]. Our SLR revealed that the CRIS-CTMS marketplace remains fragmented, largely due to the absence of comprehensive and fully integrated medical feature domains. Major AI-powered CRIS-CTMS vendors such as Saama Tech, Owkin, ConcertAI, PathAI, Lantern Pharma, Unlearn, and AiCure were included in our analysis. To enable a robust assessment, we developed a novel 360-degree maturity blueprint mapped to the GCP principles, defining essential medical feature domains for future AI-powered CRIS-CTMS implementations. Using this framework, Saama Tech demonstrated the highest level of maturity, followed by ConcertAI and Owkin. Notably, the most prominent gaps were observed in the limited integration and underdevelopment of advanced medical feature domains, as well as insufficient seamless coordination across critical clinical research activities.

The surveyed vendors incorporated specific ML and deep learning techniques such as computer vision, multimodal AI, predictive inference, federated learning, and digital twins applied to specific clinical use cases including patient recruitment, trial outcome prediction, and workflow optimization. Nevertheless, the adoption of more advanced AI and cloud-native technologies is inconsistent and typically confined to particular functions. This highlights the pressing need to prioritize the system integration of medical features within clinical research systems. Such advancements will nurture cohesive, AI-supported portfolios and mitigate the fragmented clinical solutions documented in the literature.

As digitalization and AI readiness progress in pharmaceutical and biotechnology companies, further contributions are expected to arise across the landscapes of CRIS-CTMS solutions. Advanced AI use cases, such as large language models (LLMs), MLOps, and explainable AI (XAI), are anticipated to profoundly impact clinical research information systems. Notably, most leading AI-powered CRIS-CTMS solutions examined in our SLR exhibit limitations regarding the adoption of these technologies to enhance decision support, automate complex ML workflows, and ensure transparency while building trust in data-driven processes.

The integration of LLMs [182] into CRIS-CTMS applications is expected to enhance patient matching and satisfaction through personalized interactions. Specifically, LLMs are expected to generate accurate, high-quality responses to clinical inquiries based on a deep comprehension of the medical corpus and context. This capability optimizes trial query management for clinical research associates, project managers, and medical investigators. Furthermore, LLMs based on agentic frameworks [180] will augment CRIS-CTMS solutions by incorporating reasoning agents capable of coordinating complex workflows throughout various trial phases. Tasks supported by these agents include patient recruitment, site management, data analysis, and regulatory compliance. Through dynamic data interaction and context-aware decision-making, these agents will initiate specific actions in real time, thus improving overall trial efficiency. Leveraging retrieval augmented generation (RAG) techniques [186], LLMs can access current, domain-specific knowledge such as good clinical practice. This enables them to generate accurate, timely, and citable responses that ensure adherence to evolving trial parameters and industry standards.

In parallel, MLOps [179] applied to CRIS-CTMS platforms can facilitate the detection and remediation of model and data drift [183]. This is relevant across various contexts, such as variations between patient groups or trial phases, and even across trials within the same or different therapeutic areas. MLOps enable the effective monitoring and retraining of ML models as part of the model serving process. Notably, technologies like Kubeflow [184], when incorporated into CRIS-CTMS implementations, are expected to automate model deployment, monitoring, serving, and retraining to address drift. Such capabilities are essential for maintaining predictive models’ accuracy, particularly in areas such as site management, patient recruitment, and operational execution, where data distributions may evolve over time. Continuous monitoring and retraining via MLOps ensure that trial management systems can adapt to shifts in patient populations, recruitment patterns, and site-specific factors, ultimately optimizing resource allocation and improving trial success rates.

XAI [178] will play an essential role in elucidating insights and recommendations made by a CRIS-CTMS, enabling users to make model-based, informed clinical decisions. As the need to build trust in ML models has increased, XAI can effectively support post hoc explainability and assessment methods. This facilitates engagement and collaboration with patients and regulatory bodies, which will be a key value proposition for advanced AI-powered CRIS-CTMS solutions. XAI technologies, such as Bayesian neural networks (BNNs) [181] and Shapley value-based explanations [185], will significantly enhance medical feature domains in CRIS-CTMS. BNNs provide probabilistic predictions and confidence intervals, improving decision-making in areas such as study design, data analysis, and patient recruitment by quantifying uncertainty and increasing the reliability of predictions. Shapley value-based explanations are expected to enhance transparency within CRIS-CTMS implementations. They clarify the influence of individual features on model predictions, thereby improving interpretability and fostering trust in AI-driven decision-making processes across various stages of clinical trials.

Finally, we encourage both the scientific and business communities to capitalize on the findings and contributions presented in this research. Within the domain of computer-based clinical information systems, further investigation is required to accelerate the commoditization of AI and cloud-native computing. Such progress will enable MNEs to achieve transformative outcomes and drive progress in medical treatments and pharmaceutical innovation for the broader benefit of society.