Category: Uncategorized

Introduction to AI in Healthcare

Artificial Intelligence (AI) is revolutionizing healthcare, offering unprecedented advancements in diagnostic accuracy, personalized treatment plans, and operational efficiencies. From AI-driven imaging analysis to predictive analytics, these technologies are becoming integral to modern healthcare systems. Reflecting this rapid growth, as of September 2024, the U.S. Food and Drug Administration (FDA) has authorized over 985 AI/ML-enabled medical devices, signifying a substantial increase from just a few dozen devices years prior (Figure 1). The global AI in healthcare market is projected to reach $194.4 billion by 2030, growing at a CAGR of 38.4% from 2022 to 2030.

However, AI integration presents challenges, including recalls. A 2023 study found 211 AI/ML medical device recalls between 2019 and 2021, mostly moderate risk. Examples include software errors in radiation therapy (K190387) and cardiac ultrasound (K20062). Managing these risks is crucial for AI’s future in healthcare.

Each region has its definitions and guidelines governing AI in healthcare, making compliance a critical yet intricate task for innovators. This article delves into how regulatory authorities define AI, explores global regulatory frameworks, and offers insights on successfully bringing AI innovations to market.

Product/Technology Definition

Understanding how regulatory bodies define AI is crucial for compliance and successful market entry. Below are specific definitions from key regulatory authorities:

United States (FDA):

• Artificial Intelligence is a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions influencing real or virtual environments. Artificial intelligence systems use machine- and human-based inputs to perceive real and virtual environments; abstract such perceptions into models through analysis in an automated manner; and use model inference to formulate options for information or action.

European Union (EU):

• Artificial intelligence (AI) refers to systems that display intelligent behavior by analyzing their environment and taking actions – with some degree of autonomy – to achieve specific goals. AI-based systems can be purely software-based, acting in the virtual world (e.g., voice assistants, image analysis software, search engines, speech, and face recognition systems) or AI can be embedded in hardware devices (e.g., advanced robots, autonomous cars, drones, or Internet of Things applications).

Key AI Technologies in Healthcare:

• Machine Learning is a set of techniques that can be used to train AI algorithms to improve performance at a task based on data.

• Natural Language Processing (NLP): Enables computers to understand and interpret human language, facilitating the analysis of clinical documentation.

• Computer Vision: Allows machines to interpret visual data from medical images like X-rays, MRIs, and CT scans.

Global Regulatory Frameworks for AI in Healthcare

Comparison of Global Regulatory Frameworks

Navigating the complex regulatory environment requires an understanding of how different regions approach AI in healthcare.

In the U.S., AI healthcare products are regulated by the FDA under existing medical device frameworks. For approval, your product must go through one of three main premarket pathways: 510(k), De Novo, or PMA (Pre-Market Approval). The 510(k) premarket notification is for products that are similar to an existing, already-approved device, allowing for a quicker clearance process. If your product is new, medium risk, and does not have a comparable product on the market, you will likely go through the De Novo process. For high-risk AI products, you will need to submit a premarket approval (PMA), which requires a comprehensive data set to prove the product’s safety and effectiveness.

In addition to premarket submissions, the FDA emphasizes Good Machine Learning Practices (GMLP) to ensure that your AI system follows best practices in data management, transparency, and reliability. Those practices are implemented in a Total Product Lifecycle (TPLC) approach. This means that even after your AI product is on the market, you must continuously monitor its performance, update it if necessary, and report any issues to ensure ongoing safety and effectiveness.

In the European Union, AI products used in healthcare are regulated under the Medical Device Regulation (MDR), which includes a thorough process to assess product safety and effectiveness. One of the first steps is risk classification, where you determine how risky your AI product is to patients or users. Higher-risk products undergo stricter review. After classification, the next step is the conformity assessment, where you work with a “Notified Body” (an official EU organization) to prove that your product meets the required safety and performance standards. In addition, the EU AI Act is a comprehensive regulation that will impose requirements on AI systems, particularly those deemed high-risk, across various sectors including healthcare. It complements existing regulations like the Medical Devices Regulation (MDR) and In Vitro Diagnostic Regulation (IVDR). For healthcare AI, the Act focuses on transparency, accountability, and risk management throughout the AI lifecycle. High-risk AI systems in healthcare, such as those used for emergency triage or determining eligibility for health services, will face stringent requirements. The Act aims to ensure patient safety and rights while fostering innovation in AI-enabled healthcare technologies.

Both the FDA and EMA frameworks require not only upfront proof that your AI product works but also ongoing efforts to ensure safety and effectiveness even after the product is on the market.

Case Study: FDA Approval of Viz.ai

Background

Viz.ai is an Israeli company founded in 2016 pioneering the use of artificial intelligence to accelerate the detection and treatment of strokes. The company’s AI-powered platform is designed to analyze medical imaging and automatically detect large vessel occlusion (LVO) strokes, potentially life-threatening conditions that require immediate treatment. The platform alerts stroke teams directly through their mobile devices, significantly reducing the time to intervention, which is crucial in minimizing the damage caused by strokes.

Regulatory Pathway

As can be seen in Table 2, Viz has received multiple device clearances from the FDA. Starting with the DeNovo reclassification that created new product codes, and then following with 510K submissions on modifications or new features. Viz shows us that developing a regulatory plan in parallel with your development plan is critical to releasing products to the market in an efficient manner. Note that the first clearance was only 2 years after the company was established.

Table 1 Viz FDA Regulatory Strategy Implementation Summary

A key aspect of Viz.ai’s regulatory strategy is its focus on leveraging programs that expedite market entry. For instance, Viz.ai received the FDA’s Breakthrough Device Designation, which fast-tracks the review process for technologies that provide more effective treatment or diagnosis of life-threatening conditions. Additionally, the DeNovo approval for ContaCT in 2018 (DEN170073) allowed Viz.ai to introduce a tool that identifies and communicates specific patient images to specialists, supporting decision-making in a parallel workflow without disrupting standard care.

Viz.ai supported this with a retrospective study to assess the sensitivity and specificity of ContaCT’s image analysis and notification system, using 300 CT angiogram studies from two U.S. clinical sites for comparison against neuro-radiologist assessments.

This early achievement paved the way for ongoing improvements and feature expansions through subsequent 510(k) submissions, as outlined in the table above. These regulatory advancements allowed Viz.ai to continually refine and enhance its technology, ensuring it remains at the forefront of innovation in patient care and diagnostic efficiency.

Compliance with FDA Definitions and Requirements

• Alignment with FDA’s AI Definition: The Viz.ai platform aligns with the FDA’s definition of AI-based medical devices. It uses AI algorithms to assist in the real-time analysis of CT scans and other medical imaging, mimicking cognitive functions related to human intelligence in the decision-making process.

• Safety & Efficacy: Viz.ai provided extensive clinical performance data showing a reduction in time from image acquisition to stroke team notification by 52 minutes, significantly improving patient outcomes.

• This real-time analysis and alert system help reduce the time to treatment, which is crucial as “time is brain” in stroke management.

Outcome

• FDA Clearance: Viz.ai’s LVO Stroke Platform became the first FDA-approved AI platform for stroke detection. The platform is now used in over 1,000 hospitals across the U.S., Israel, and Europe.

• Impact: By reducing the time to treatment, Viz.ai has significantly improved stroke care, providing earlier intervention, which leads to better patient outcomes and reduced long-term disability rates. The AI-powered alert system ensures that stroke teams are notified promptly, saving critical time in decision-making and patient care.

Lessons Learned

• Innovation and Collaboration: Viz.ai’s success shows how combining innovative AI technologies with ongoing and early regulatory engagement can bring transformative healthcare solutions to market quickly.

• Proactive Regulatory Approach: By engaging with the FDA’s Breakthrough Devices Program early, Viz.ai was able to expedite the regulatory process, ensuring faster market approval.

• Real-World Impact: The success of Viz.ai demonstrates that AI can revolutionize the speed and effectiveness of treatments for time-critical conditions, such as stroke, ultimately saving lives.

Conclusion

The integration of AI into healthcare is revolutionizing the field, offering groundbreaking advancements in diagnostics, treatment, and operational efficiency. However, navigating the complex global regulatory landscape is essential for these technologies to succeed. Different regulatory bodies, such as the FDA and EMA, provide specific frameworks that innovators must align with, ensuring safety, effectiveness, and compliance throughout the product lifecycle. Early regulatory engagement, as shown in case studies, can expedite approval processes, while ongoing compliance with data privacy and safety standards remains crucial. Companies that prioritize regulatory strategy are better positioned to bring innovations to market swiftly and safely, driving progress in this transformative era of healthcare.

Stay tuned for upcoming editions of ‘Innovation Meets Regulation,’ where we explore the intersection of healthcare innovation and regulatory frameworks that shape the industry’s future.

Bringing a new drug or biologic to U.S market is a complex journey that requires strict compliance with the FDA’s regulations. One key aspect of this process is the Bioresearch Monitoring (BIMO) program, which assures the quality and integrity of clinical trial data.

In December 2024, the FDA released new guidance, Standardized Format for Electronic Submission of NDA and BLA Content for the Planning of BIMO Inspections. This guidance aims to streamline submissions for New Drug Applications (NDAs) and Biologics License Applications (BLAs).

This article provides an overview of the guidance, its goals, and how it impacts companies preparing submissions. Whether you’re new to regulatory processes or an experienced professional, this guide will clarify what you need to know and do.

What is the FDA BIMO Guidance and Inspection Process?

The Bioresearch Monitoring (BIMO) program is a comprehensive FDA initiative designed to assure that clinical trials are conducted ethically and that their data is trustworthy. Here’s what it entails:

• Data Integrity: The program verifies that clinical trial data submitted to the FDA is reliable and accurately reflects study results.
• Ethical Compliance Assurance: It ensures that trials comply with Good Clinical Practices (GCP) and that participants’ rights and safety are prioritized. This sustained process helps maintain trust in trial conduct.
• BIMO Inspections: The FDA conducts on-site inspections, data audits, and remote regulatory assessments of clinical trial sites, sponsors, contract research organizations (CROs), and institutional review boards (IRBs) involved in the studies. These inspections often follow a risk-based approach to target areas of concern.

BIMO findings provide valuable information to the FDA regarding the integrity and reliability of the clinical data submitted in NDAs and BLAs. This ensures that the data is adequate to support the approval of new treatments. Without this comprehensive program, patient safety and public trust in the drug approval process could be at risk.

The New Guidance for BIMO Inspections: Aim and Main Changes

Aim of the FDA BIMO Guidance

The new guidance aims to standardize and simplify data submission required for planning BIMO inspections. This standardization facilitates the FDA’s ability to identify sites for inspection, evaluate study integrity, and maintain an efficient review timeline.

Standard Data to Submit as Part of NDAs and BLAs

Study-specific data (both clinical study-level information and clinical site data) should be submitted at several levels:

Clinical Study-Level Information: Sponsors must provide comprehensive tables listing all clinical sites, including:

a. Investigator names, and complete site details  

b. Details of CROs/vendors or other entities to whom the Sponsor delegated activities.

c. Protocol (and amendments) and Annotated CRF

Subject-Level Data Line Listings: Detailed, site-specific data for each trial participant must be provided, including studies with different treatment indications. Primary data points (in addition to derived data) must be submitted. For example, in a pain trial, raw diary entries of participants’ pain scores (primary data) must be submitted alongside calculated results (derived data).

Summary-Level Clinical Site Dataset: This dataset summarizes key information from clinical sites across all pivotal studies. It allows the FDA to effectively prioritize sites for inspection.

All submissions must adhere to the FDA’s electronic common technical document (eCTD) format and be uploaded through the FDA’s secure gateway.

Key Components of Data Standardization of Submission Content – FDA BIMO Inspection Guidance

What Does This Mean for Your Business?

These changes streamline regulatory submissions and improve the efficiency of FDA reviews. However, they also require more detailed preparation and coordination. Sponsors should consider the following actionable steps:

1. Streamlined Submissions: Ensure data formats are standardized to reduce variability and errors. This will enable smoother FDA reviews and minimize back-and-forth requests for clarification.
2. Risk-Based Inspections: Be aware that the FDA employs a risk-based model to prioritize site inspections, meaning sites with incomplete or inconsistent data may face closer scrutiny. Proactively address potential gaps in your data.
3. Early Integration: Incorporate the guidance’s requirements into clinical trial planning from the outset to avoid delays during submission. Early compliance ensures smoother regulatory processes.

Steps to Stay Ahead

1. Prepare for Raw Data Requirements: Configure your EDC system to capture detailed raw data, not just calculated scores. Implement edit checks to verify data accuracy and plan for source data verification (SDV) for critical measurements.
2. Enhance Risk Management: Assess how these requirements may impact your risk management processes and address vulnerabilities early.
3. Foster Targeted Collaboration: Align regulatory, clinical, and data management teams to address the specific demands of the new requirements. Early coordination is key to streamlining processes and avoiding delays.
4. Enhance Team Readiness: Sensitize stakeholders, including investigators, CROs, and vendors, about the requirements and their role in providing data to the FDA. Enhanced scrutiny may increase the likelihood of regulatory inspections.

Conclusion

The FDA’s updated BIMO guidance and inspection processes reflect its efforts to prioritize efficiency in regulatory processes while enabling enhanced scrutiny of sponsors, investigators, and CROs. By adopting the standardized format for data submission, companies can expedite their review timelines and demonstrate a commitment to transparency and data integrity.

Join us for Digital Health: The FDA Insider’s View, an exclusive webinar offering key insights into best practices for healthcare technology, focusing on AI/ML, FDA cybersecurity expectations, and software documentation. Learn firsthand how to navigate regulatory submissions and clinical trials.