jrnls.org logo
Submit

BROWSE ALL JOURNALS

SEE ALL 19 JOURNALS
International Journal of Pharmaceutical InvestigationNOW IN
Review

A Global Perspective on Navigating the Regulatory Landscape of Adaptive AI in Healthcare: Balancing Innovation and Patient Safety

Sandeep Talari1, Vivek Reddy Murthannagari2, Gundavarapu Kanaka Phani Kumar3, Gowthamarajan Kuppusamy4, Srikanth Jupudi1★★ Corresponding author

  1. 1Department of Pharmaceutical Chemistry, JSS College of Pharmacy, JSS Academy of Higher Education and Research, Ooty, Nilgiris, Tamil Nadu, INDIA.
  2. 2Department of Pharmaceutical Regulatory Affairs, JSS College of Pharmacy, JSS Academy of Higher Education and Research, Ooty, Nilgiris, Tamil Nadu, INDIA.
  3. 3Pharmaceutical Development, Aurigene Oncology Limited, Bengaluru, Karnataka, INDIA.
  4. 4Department of Pharmaceutics, JSS College of Pharmacy, JSS Academy of Higher Education and Research, Ooty, Nilgiris, Tamil Nadu, INDIA.

CORRESPONDENCE

Srikanth Jupudi

Assistant Professor, Department of Pharmaceutical Chemistry, JSS College of Pharmacy, JSS Academy of Higher Education and Research, Ooty-643001, Nilgiris, Tamil Nadu, INDIA.

sjphd@jssuni.edu.in

Received: 02-06-2025; Revised: 31-07-2025; Accepted: 15-09-2025.

Volume 16, Issue 1 · pp. 23–29 · PUBLISHED Jan-Mar 2026 · DOI: 10.5530/ijpi.20260314

View on IJPI original site ↗

ABSTRACT

In healthcare, adaptive AI and machine learning offer continuous performance improvement through patient data updates, but traditional regulation methods are slow and cumbersome. The USFDA and EMEA propose a paradigm shift, focusing on oversight of quality management systems and algorithm change protocols to ensure patient safety while enabling advancements. Regulatory challenges arise from the need for frequent updates, which require a robust and efficient method. This article explores the regulatory landscape for AI-based medical devices highlighting efforts by Japan, the EU, and the US, and the perspectives of organizations like IMDRF and WHO.

KEYWORDS

0% READ

FULL TEXT

INTRODUCTION

Adaptive artificial intelligence and machine learning are promising for healthcare by enabling continuous improvement through patient data analysis. However, current regulatory methods, such as those from the FDA and EMA, struggle with the speed of updates, requiring extensive documentation and revalidation. To address this, new oversight frameworks aim to balance innovation with patient safety. The article evaluates the regulatory challenges and approaches taken by the U.S., Europe, and Japan to ensure the efficacy and safety of adaptive AI devices. It reviews the regulatory frameworks and methodologies for assessing the safety and effectiveness of Software as a Medical Device (SaMD) (Candelon et al., 2021).

The advent of AI - based technology represents an ongoing fourth industrial revolution, it has brought with it novel issues concerning the regulatory landscape for such innovative products that have yet to be fully comprehended (Hosoda et al., 2020; Hosoda et al., 2020; Stead, 2018; Topol, 2019; Verghese et al., 2018; Aisu et al., 2022). This poses a challenge for global regulatory bodies like the USFDA (USA), EMA (EU), and PMDA (Japan) due to varying needs, socioeconomic factors, and cultural nuances across regions, complicating regulatory harmonization (Reddy, 2024). Collecting, processing, and sharing personal information presents ethical concerns regarding data security and the risk of online attacks. Accountability is important if decisions lead to negative consequences (Cohen et al., 2020), and transparency is necessary because decision tree data is often stored in an inaccessible 'Blackbox'. Aristotle valued reproducibility and empirical verification of the methods of achieving results over the need for explainability, further, bias is also significant impediment while evaluating societal issues (Coeckelbergh, 2019). The regulation of AI-based devices poses challenges for policymakers in making an overarching policy, for instance, AI mammogram analysis requires diverse data sets to address racial differences such as breast density variations (Babic et al., 2019; Cohen et al., 2020). Organizations like the International Medical Device Regulators Forum are at the forefront of efforts to harmonize these standards.

Some commonly used SaMD includes the AI-enabled application Philips’ IntelliVue Guardian solution (Intellivue, n.d.) (interfaces with blood pressure monitors, oximeters, and ECGs to collect data, predict patterns, and issue alerts before emergencies), the Apple ECG app and Apple Irregular Rhythm notifications, (Benjamens et al., 2020; U.S. Food and Drug Administration, 2018) the Glooko’s Diabetes Management Platform (Doyle-Delgado and Chamberlain, 2020; Glooko, 2025), and the Ginger’s Mental Health Care Platform (Ginger, 2025). AiCure is a drug adherence monitoring SaMD which improves medication adherence, in the U.S., medical advice is often ignored, leading to deaths and financial loss when it can be prevented with this aid (Babel et al., 2021; Ebner et al., 2016; AiCure, 2025). Technologies like ingestible sensors with RFID tags and smart pill dispensers can help track adherence anywhere and anytime (Schreier, 2014; Ebner et al., 2016). Nevertheless, the long-term effects of ingestible sensors on patients must be considered.

INTEGRATING AI INTO HEALTHCARE

AI enables devices to predict outcomes using algorithms and mimic human actions (European Commission, 2020). In the pharmaceutical industry, AI solutions are gaining traction for managing vast health data efficiently-from drug discovery to commercialization-thus saving time and money. AI converts unstructured medical data into actionable insights, reduces biases, and automates repetitive tasks, creating increased industry value (Johnson et al., 2018).

There are three categories in which software can be integrated with a medical device and the best way to distinguish between the three software categories is through the terminology used. SiMD is software integrated into a medical device to enhance its performance. In contrast, SaMD functions independently as a medical device. MDSW is a term used in the EU regulatory framework for software intended for use alone or with other software, as specified in the medical device’s regulation or in vitro diagnostic medical devices regulation (Spanou D, 2013).

REGIONAL REGULATORY FRAMEWORKS

Japan: Regulatory Framework For Updating SaMD

Medical devices in Japan are reviewed by PMDA and approved by MHLW. Japan follows IMDRF risk classifications viz: Class 1, Class 2, Class 3, and Class 4. Class 1 devices need notification or self-certification because of the minimal or no risk to patient safety. Class 2 devices need third-party certification. Class 3 and above need review and approval (Aisu et al., 2022). If the certification standards are established, there is no need for review. When it comes to adaptive SaMD, Japan's regulatory framework requires approval for SaMD because of its adaptive nature.

The successful deployment of adaptive SaMD in the market hinges on the quality of input data, as the system learns and evolves through exposure to new patients. To ensure the efficacy and safety of these modified MDs, both the MHLW and PMDA mandate submission of clinical data and post-marketing change plans (partial and minor) for application review. It is worth noting that the MHLW may approve change applications not solely by reviewing clinical data, but also by verifying whether the changes align with the change plan (Ota et al., 2020). It is expected that this approach expedites the process and allows for more timely access to modified MDs for patients. The quality of input data is of paramount importance in frequently updating AI MD. In Japan, there is a well-established system to collect health data from individuals throughout their lives, starting from before birth. This system was set up under the Next Generation Medical Infrastructure Act, which became effective in 2018. With this data available, there are ample opportunities for academic institutions, government bodies, commercial enterprises, and businesses to conduct research. Notably, there are no concerns regarding the privacy and safe use and retention of personal data of individuals (Ota et al., 2020; Yamamoto, 2022). This feature presents a significant advantage if used responsibly. The regulatory process steps for an adaptive SaMD was illustrated in Figure 1.

Figure 1: The regulatory process steps for an adaptive SaMD (Isakadze and Martin, 2020).

EUROPEAN UNION COMMISSION DIRECTIVES

The EU Commission has collaborated with the MDCG to regulate SaMD under the EU Medical Device Regulation (MDR; 2017/745) and IVDR (2017/746) (European Medicines Agency, n.d.), focusing on post-marketing surveillance and the EUDAMED database (Chhaya and Khambholig, 2021). A regulatory framework for AI has been proposed, in the view of categorizing risks into four levels: unacceptable risk, high risk, limited risk, and minimal or no risk (Mason Hayes and Curran, n.d.; European Commission, 2020). High-risk AI will be prioritized, and unacceptable risk AI will be removed from the market, aligning with the MDR classification in article 2(1) for MDSW.

The EU’s drafted legislation that mandates the assignment of EU product conformity and CE marking concepts to all "high-risk" AI applications. Notably, AI in medical devices is deemed high-risk by default. The drafted legislation bears considerable resemblance to medical device regulation concerning core approaches, such as product intended use, post-market monitoring systems, and notified bodies (Figure 2) (Decos, 2023).

Figure 2: The process steps for evaluating high-risk AI model (Sierra Labs, n.d.; Lubbers et al., 2021).

The EU-specific solutions that the draft legislation calls for are particularly necessary in three domains. Firstly, EU data protection considerations related to the update problem. Secondly, the relatively less established system for RWP monitoring. Thirdly, the EU's public perceptions concerning the role of AI (Gilbert et al., 2021; Petersen et al., 2021). It is noteworthy that, in the EU's healthcare sector, other factors such as EU core values, political influence, and global competition are of paramount importance and must be duly considered. The draft legislation highlights potential trade-offs, such as those between investment and transparency (black box concept), explainability and accuracy, and accountability (Cohen et al., 2020). Also, the path ahead involves collaborating with SMEs in the field and maintaining a complete life cycle of AI systems, including establishing quality data, risk management, and final withdrawal.

THE USFDA

Medical devices in the U.S. are regulated by the CDRH within the FDA. CDRH oversees device regulation, including approval and post-market surveillance. The FDA classifies devices into three categories: Class I (minimum risk), Class II (medium risk), and Class III (high risk) based on IMDRF guidelines. The current approval process for SaMD is similar to the 510(k)-notification process. Low-risk devices are assumed safe and must follow Good Manufacturing Practices. The FDA primarily uses the 510(k) pathway if there’s an equivalent predicate device. Manufacturers need to prove their product's equivalence for approval. If no predicate exists, sponsors can file for PMA or De Novo for entirely new low to moderate-risk devices (U.S. Food and Drug Administration, 2019). The approval process varies by risk classification (Table 1).

Table 1: The medical device regulations in the 800 series of the 21st Code of Federal Regulation cover general provisions, labeling, quality systems, and post-market surveillance.
PartTitlepartTitle
800General Provisions801Labeling Requirements for Unique Device Identification.
803Medical Device Reporting806Reports of Corrections and Removals.
807Establishment Registration and Device Listing.810Medical Device Recall Authority.
809In Vitro Diagnostic Products for Human Use.814Premarket Approval of Medical Devices.
812Investigational Device Exemptions821Medical Device Tracking Requirements.
820Quality System Regulation830Unique Device Identification
822Post-market Surveillance861Procedures for Performance Standards Development.
860Medical Device Classification Procedures.862Clinical Chemistry and Clinical Toxicology Devices.

The USFDA Proposed a Framework for Modifications to Adaptive SaMD

The US FDA has approved several SaMD products that use AI and ML algorithms, typically requiring these algorithms to remain unchanged before marketing. If modifications extend beyond the original authorization, FDA premarket review is needed. However, some algorithms can adapt and optimize based on real-world experiences, potentially producing different outputs from those initially approved.

Ensuring device safety is crucial after algorithm changes, as these can either enhance or compromise performance. To address this, the US FDA recommends a TPLC approach (U.S. Food and Drug Administration, 2019). In addition, it's important to set boundaries for the evolving scope of AI/ML-based SaMDs. This includes defining SPS, establishing an ACP, following GMLP, ensuring transparency, and conducting real-world performance monitoring. These elements help ensure continued safety and efficacy amid algorithm modifications.

The PCCP: The USFDA has established a review and approval process for the SPS and ACP, with each device being expected to include a two-component PCCP. The SPS outlines permissible changes to the AI, while the ACP provide a process to ensure modifications maintain safety and effectiveness. Real-world performance monitoring is necessary to confirm ongoing safety and efficacy, while any changes outside the PCCP require a regulatory review. Feedback from various stakeholders has resulted in a 5-point action plan designed to address concerns and enhance the proposed framework (U.S. Food and Drug Administration, 2024; U.S. Food and Drug Administration, 2018).

The Final PCCP guidance issued in December 2024 clarifies regulatory expectations for manufacturers, updating terminology to AI-enabled Device Software Function (AI-DSF) and focusing on device-led combination products. It outlines submission types such as Original PMA, Modular PMA, and 510(k) submissions, and emphasizes the importance of version control and maintenance to ensure changes are managed effectively. The guidance also details an interactive review process, allowing the FDA to request further information if necessary (U.S. Food and Drug Administration, 2024). These updates aim to support innovation in AI technologies while ensuring patient safety and device effectiveness.

Major regulators are also adopting international standards such as the software life cycle, software usability testing, validation, release, and update along with their attempts to regulate frequently modifying AI Medical Devices (Babic et al., 2019). This article provides an overview of the most important standards related to the development of SaMD in the coming section.

A Glance at Software Compliance Standards for SaMD

The term "standard" refers to requirements and specifications that ensure materials, products, processes, and services are fit for their intended purpose. The rise in software use in the medical industry has led to frameworks for its development. To ensure medical device software meets these standards, the ISO and the IEC provide guidelines on development, maintenance, and best practices (Table 2) (Decos, 2023; Stead, 2018; Topol, 2019; U.S. Food and Drug Administration, 2018).

Table 2: List of applicable standards concerning Software as Medical Devices (Decos, 2023; Stead, 2018; Topol, 2019; U.S. Food and Drug Administration, 2018).
StandardDescription
ISO 13485To ensure the quality and safety of SaMD, a QMS incorporates ISO 13485 for overall quality management, IEC 62304 for software life cycle processes, ISO 14971 for risk management, and IEC 62366 for usability engineering; this integrated approach addresses design, risk, usability, and maintenance throughout the SaMD lifecycle.
IEC 62304
ISO 14971
IEC 62366
IEC 81001This standard focuses on cybersecurity in medical devices, including SaMD. It outlines the necessary security measures throughout the software lifecycle, ensuring that SaMD is secure from development to maintenance.
IEC 60601While primarily for electrical medical equipment, this standard also includes guidelines for programmable electrical medical systems, which can encompass SaMD.
IEC 82304This is the specific standard for SaMD, covering general requirements for the safety and security of health software products.
ISO 10993Although this standard is about biocompatibility, it's relevant if the SaMD interacts with patients at the cellular or bodily fluid levels.

Developers should follow four standards that constitute a quality management system for full lifecycle traceability and broad acceptance. However, ISO and IEC have not adequately addressed procedures for managing adaptive SaMD.

DISCUSSION

Major regulators are leveraging existing principles to review and approve adaptive AI Medical Devices, saving time and resources (Table 3). The regulatory frameworks for SaMD in Japan, the EU, and the US show both similarities and differences. All three regions use a risk-based classification system and emphasize post-market surveillance and continuous monitoring of AI/ML-based medical devices. The EU's detailed four-level AI risk categorization (Mason Hayes and Curran, n.d.; European Commission, 2020) contrasts with the US and Japan. Each has defined pathways for SaMD approval, like PMDA reviews in Japan, MDR and IVDR frameworks in the EU, and the 510(k), PMA, and De Novo pathways in the US. They focus on maintaining high standards of quality, safety, and effectiveness, requiring clinical data and real-world performance monitoring. However, the EU emphasizes data protection under GDPR (Gilbert et al., 2021), whereas Japan, with its Next Generation Medical Infrastructure Act, has a more flexible approach (Ota et al., 2020; Yamamoto, 2022). The US has HIPAA but it's less comprehensive compared to GDPR (U.S. Food and Drug Administration, 2019). Regulatory processes also differ with Japan's PMDA and MHLW, the EU's centralized MDCG (Aisu et al., 2022; Chhaya and Khambholig, 2021), and the US FDA's CDRH (U.S. Food and Drug Administration, 2019; U.S. Food and Drug Administration, 2018).

Table 3: Comparison of Japan, EU, and US Frameworks for Adaptive SaMD.
AspectJapanEUUS FDA
Regulations-Next Generation Medical Infrastructure Act -Oversight by PMDA and MHLW.-MDR (2017/745) -IVDR (2017/746) -Proposed AI regulatory framework.-Oversight by CDRH within the FDA -21st Century Cures Act.
Risk ClassificationFour- level risk classifications.Four-level AI risk categorization.Three-level AI risk categorization.
SaMD ClassificationAdaptive SaMD requires premarket approval.Conformity and CE marking to all ‘high-risk’ AI application.SaMD follows similar pathways as other medical devices (510(k), PMA, De Novo).
Pre-Market Requirements-Clinical data submission. -Post-marketing change plans.-Conformity assessment -Notified Bodies for high-risk devices. -Technical documentation.-510(k) for devices with predicates -PMA or De Novo for novel devices.
Adaptive SaMD Modifications-MHLW approval based on alignment with pre-defined change plans.-Generally requires new conformity assessment if changes impact intended use or safety.-Total Product Lifecycle approach -PCCP, including SaMD Pre-Specifications and Algorithm Change Protocol.
Post-Market SurveillanceRequired, with emphasis on change management.Required, EUDAMED is the central database used for enhanced post-market surveillance.Required, with Real-World Performance monitoring (RWP).
Data UsageEstablished system for lifelong health data collection.Stringent GDPR requirements.Focus on data integrity and patient privacy under HIPAA.
Other ConsiderationsExpedited approvals, focus on input data quality.EU core values, political influence, global competition, SME collaboration.Good Machine Learning Practices, transparency, PCCP review process (Original PMA, Modular PMA, 510(k)), version control.

The ISO and IEC standards are internationally recognized and widely adopted by major regulatory (U.S. Food and Drug Administration, 2019) bodies such as the USFDA, EMA, and PMDA. These standards form the basis for quality management, risk management, software life cycle processes, and usability engineering in medical devices, ensuring safety, effectiveness, and compliance across global markets (NQA, n.d.; IEC 62366, 2023).

The future prospects might include harmonizing regulations through bodies like IMDRF and WHO (World Health Organization, 2021), evolving AI regulations (U.S. Food and Drug Administration, 2019), securing data utilization, especially in Japan, and building public trust through transparency and stakeholder involvement (Gilbert et al., 2021)

CONCLUSION

A SaMD's behavior may change with real-world data from the next patient encounter, making conventional methods laborious. Regulators are considering 'predefined change plans' to be submitted as part of the submission to manage such adaptations. However, not all valuable modifications can be anticipated from the outset. Harmonizing regulatory requirements is complex due to varying data quality, conditions, socioeconomic, cultural, and epidemiological factors, and legal and ethical considerations, in each region. To address regional differences, establish global considerations with added local specifics for SaMD. This approach ensures the global framework adapts to local nuances, enhancing the effectiveness and safety of SaMD across regions.

REFERENCES

As published

Showing references and in-text citations exactly as published.

  1. 1.AiCure.. Medication adherence. n.d..GOOGLE SCHOLAR
  2. 2.Aisu, N.; Miyake, M.; Takeshita, K.; Akiyama, M.; Kawasaki, R.; Kashiwagi, K. et al. Regulatory-approved deep learning/machine learning-based medical devices in Japan as of 2020: A systematic review. PLOS Digital Health. 2022;1(1):Article e0000001. https://doi.org/10.1371/journal.pdig.0000001DOIGOOGLE SCHOLAR
  3. 3.Babel, A.; Taneja, R.; Mondello Malvestiti, F.; Monaco, A.; Donde, S.. Artificial intelligence solutions to increase medication adherence in patients with non-communicable diseases. Frontiers in Digital Health. 2021;3:Article 669869. https://doi.org/10.3389/fdgth.2021.669869DOIGOOGLE SCHOLAR
  4. 4.Babic, B.; Gerke, S.; Evgeniou, T.; Cohen, I. G.. Algorithms on regulatory lockdown in medicine. Science. 2019;366(6470):1202–1204. https://doi.org/10.1126/science.aay9547DOIGOOGLE SCHOLAR
  5. 5.Benjamens, S.; Dhunnoo, P.; Meskó, B.. The state of artificial intelligence- based FDA-approved medical devices and algorithms: An online database. npj Digital Medicine. 2020;3(1):118. https://doi.org/10.1038/s41746-020-00324-0DOIGOOGLE SCHOLAR
  6. 6.Candelon, F.; Di Carlo, R. C.; De Bondt, M.; Evgeniou, T.. AI regulation is coming. Harvard Business Review. 2021;99(5):103–111.GOOGLE SCHOLAR
  7. 7.Chhaya, V.; Khambholig, K.. The SaMD regulatory landscape in the US and Europe. 2021.GOOGLE SCHOLAR
  8. 8.Clark, P.; Kim, J.; Aphinyanaphongs, Y.. Marketing and US Food and Drug Administration clearance of artificial intelligence and machine learning enabled software in and as medical devices: A systematic review. JAMA Network Open. 2023;6(7):Article e2321792. https://doi.org/10.1001/jamanetworkopen.2023.21792DOIGOOGLE SCHOLAR
  9. 9.Coeckelbergh, M.. Artificial intelligence: Some ethical issues and regulatory challenges. Technology and Regulation. 2019;2019:31–34. https://doi.org/10.71265/a9yxhg88DOIGOOGLE SCHOLAR
  10. 10.Cohen, I. G.; Evgeniou, T.; Gerke, S.; Minssen, T.. The European artificial intelligence strategy: Implications and challenges for digital health. The Lancet. Digital Health. 2020;2(7):e376–e379. https://doi.org/10.1016/S2589-7500(20DOIGOOGLE SCHOLAR
  11. 11.Decos. Software As a Medical Device (SAMD) in EU MDR and its submission process. 2023.GOOGLE SCHOLAR
  12. 12.Doyle-Delgado, K.; Chamberlain, J. J.. Use of diabetes-related applications and digital health tools by people with diabetes and their health care providers. Clinical Diabetes. 2020;38(5):449–461. https://doi.org/10.2337/cd20-0046DOIGOOGLE SCHOLAR
  13. 13.Ebner, H.; Hayn, D.; Kropf, M.; Modre-Osprian, R.; Schreier, G.. Internet of things based medication adherence assessment. In Studies in Health Technology and Informatics. IOS Press. 2016;221:(118–118. https://doi.org/10.3233/978-1-61499-633-0-118DOIGOOGLE SCHOLAR
  14. 14.European Commission.. White paper on artificial intelligence-a European approach to excellence and trust. Europa. 2020.GOOGLE SCHOLAR
  15. 15.European Medicines Agency.. Medical device legislation. n.d..GOOGLE SCHOLAR
  16. 16.Gilbert, S.; Fenech, M.; Hirsch, M.; Upadhyay, S.; Biasiucci, A.; Starlinger, J.. Algorithm change protocols in the regulation of adaptive machine learning-based medical devices. Journal of Medical Internet Research. 2021;23(10):Article e30545. https://doi.org/10.2196/30545DOIGOOGLE SCHOLAR
  17. 17.Ginger.. On-demand mental healthcare. n.d..GOOGLE SCHOLAR
  18. 18.Glooko.. Diabetes Management Platform. n.d..GOOGLE SCHOLAR
  19. 19.Guardian, I.V. [Software].. Philips. n.d..GOOGLE SCHOLAR
  20. 20.Hayes, M.; Curran. EU AI act: Risk categories. n.d..GOOGLE SCHOLAR
  21. 21.Hosoda, Y.; Miyake, M.; Meguro, A.; Tabara, Y.; Iwai, S.; Ueda-Arakawa, N. et al. Keratoconus-susceptibility gene identification by corneal thickness genome-wide association study and artificial intelligence IBM Watson. Communications Biology. 2020;3(1):410. https://doi.org/10.1038/s42003-020-01137-3DOIGOOGLE SCHOLAR
  22. 22.Hosoda, Y.; Miyake, M.; Yamashiro, K.; Ooto, S.; Takahashi, A.; Oishi, A. et al. Deep phenotype unsupervised machine learning revealed the significance of pachychoroid features in etiology and visual prognosis of age-related macular degeneration. Scientific Reports. 2020;10(1):Article 18423. https://doi.org/10.1038/s41598-020-75451-5DOIGOOGLE SCHOLAR
  23. 23.IEC 62366.. Wikipedia. 2023.GOOGLE SCHOLAR
  24. 24.Isakadze, N.; Martin, S. S.. How useful is the smartwatch ECG?. Trends in Cardiovascular Medicine. 2020;30(7):442–448. https://doi.org/10.1016/j.tcm.2019.10.010DOIGOOGLE SCHOLAR
  25. 25.Johnson, K. W.; Torres Soto, J.; Glicksberg, B. S.; Shameer, K.; Miotto, R.; Ali, M. et al. Artificial intelligence in cardiology. Journal of the American College of Cardiology. 2018;71(23):2668–2679. https://doi.org/10.1016/j.jacc.2018.03.521DOIGOOGLE SCHOLAR
  26. 26.Lubbers, B. R.; Schilhabel, A.; Cobbaert, C. M.; Gonzalez, D.; Dombrink, I.; Brüggemann, M. et al. The new EU regulation on in vitro diagnostic medical devices: Implications and preparatory actions for diagnostic laboratories. HemaSphere. 2021;5(5):e568. https://doi.org/10.1097/HS9.0000000000000568DOIGOOGLE SCHOLAR
  27. 27.NQA.. ISO 13485: 2016. Medical devices management system implementation guide. n.d..GOOGLE SCHOLAR
  28. 28.Ota, N.; Tachibana, K.; Kusakabe, T.; Sanada, S.; Kondoh, M.. A concept for a Japanese regulatory framework for emerging medical devices with frequently modified behavior. Clinical and Translational Science. 2020;13(5):877–879. https://doi.org/10.1111/cts.12784DOIGOOGLE SCHOLAR
  29. 29.Petersen, C.; Smith, J.; Freimuth, R. R.; Goodman, K. W.; Jackson, G. P.; Kannry, J. et al. Recommendations for the safe, effective use of adaptive CDS in the US healthcare system: An AMIA position paper. Journal of the American Medical Informatics Association. 2021;28(4):677–684. https://doi.org/10.1093/jamia/ocaa319DOIGOOGLE SCHOLAR
  30. 30.Reddy, S.. Global harmonisation of AI-enabled software as a medical device regulation: Addressing challenges and unifying standards. Mayo Clinic Proceedings. Digital Health. 2024.GOOGLE SCHOLAR
  31. 31.Schreier, G.. The internet of things for personalized health. pHealth. Studies in Health Technology and Informatics. 2014;200:22–31.GOOGLE SCHOLAR
  32. 32.Sierra labs.. A Simple Breakdown of ISO and IEC Standards for SaMD [Blog]. sierralabs.com. n.d..GOOGLE SCHOLAR
  33. 33.Spanou, D.. Software as a medical device (SaMD): Key definitions. International Medical Device Regulators Forum (IMDRF). 2013.GOOGLE SCHOLAR
  34. 34.Stead, W. W.. Clinical implications and challenges of artificial intelligence and deep learning. JAMA. 2018;320(11):1107–1108. https://doi.org/10.1001/jama.2018.11029DOIGOOGLE SCHOLAR
  35. 35.Topol, E. J.. High-performance medicine: The convergence of human and artificial intelligence. Nature Medicine. 2019;25(1):44–56. https://doi.org/10.1038/s41591-018-0300-7DOIGOOGLE SCHOLAR
  36. 36.United States Food and Drug Administration.. Medical device safety action plan: Protecting patients, promoting public health. 2018.GOOGLE SCHOLAR
  37. 37.United States Food and Drug Administration.. Proposed regulatory framework for modifications to Artificial Intelligence/Machine Learning (AI/ML)-based software as a medical device (SaMD). 2019.GOOGLE SCHOLAR
  38. 38.United States Food and Drug Administration.. Quality System (QS) regulation/medical device current good manufacturing practices (CGMP). 2024a.GOOGLE SCHOLAR
  39. 39.United States Food and Drug Administration.. Marketing submission recommendations for a predetermined change control plan for artificial intelligence-enabled device software functions. 2024b.GOOGLE SCHOLAR
  40. 40.Verghese, A.; Shah, N. H.; Harrington, R. A.. What this computer needs is a physician: Humanism and artificial intelligence. JAMA. 2018;319(1):19–20. https://doi.org/10.1001/jama.2017.19198DOIGOOGLE SCHOLAR
  41. 41.World Health Organization.. Guidance. Ethics and Governance of Artificial Intelligence for Health. 2021.GOOGLE SCHOLAR
  42. 42.Yamamoto, R.. The next-generation medical infrastructure law. In. Epidemiologic research on real-world medical data in Japan. Springer Nature. 2022;2:115–120. https://doi.org/10.1007/978-981-19-1622-9_15DOIGOOGLE SCHOLAR

Cite this article

SELECT FORMAT

Talari, S., Murthannagari, V. R., Kumar, G. K. P., Kuppusamy, G., & Jupudi, S. (2026). A Global Perspective on Navigating the Regulatory Landscape of Adaptive AI in Healthcare: Balancing Innovation and Patient Safety. International Journal of Pharmaceutical Investigation, 16(1), 23–29. https://doi.org/10.5530/ijpi.20260314