The FDA had issued a final rule to regulate Laboratory Developed Tests (LDTs) that was intended to fundamentally redefine the oversight field for diagnostic technologies. However, this rule was subsequently vacated by a federal court and then officially rescinded by the FDA. While the rule is no longer in effect, the discussions around this regulatory shift, which aimed to extend beyond traditional lab assays and reach deep into the architecture of clinical AI platforms, particularly those integrated into laboratory workflows, highlight the complex intersection of lab oversight and software validation. Federal healthcare regulators and diagnostics policy experts continue to consider the need for a coordinated policy response to prevent overlapping and potentially contradictory requirements in this evolving area.
The FDA’s Redrawing of Regulatory Boundaries for Diagnostic Software
The FDA’s final rule, “Medical Devices. Laboratory Developed Tests” FDA Federal Register notice for LDT proposed rule, issued on May 6, 2024, had signaled a significant departure from decades of enforcement discretion for LDTs. Historically, LDTs, which are tests designed, manufactured, and used within a single laboratory, have largely operated outside the direct purview of FDA premarket review. The agency’s rationale for this change centered on the increasing complexity and risk profile of modern LDTs, many of which now resemble commercially manufactured diagnostic kits. This reclassification would have meant that LDTs, including those incorporating advanced computational analysis, would be subject to the same regulatory requirements as other medical devices, including premarket review, quality system regulations (QSR), adverse event reporting, and labeling requirements. The rule was projected to affect nearly 80,000 existing tests offered by almost 1,200 laboratories. However, on March 31, 2025, a federal district court vacated this final rule, and on September 19, 2025, the FDA officially rescinded it, reverting to its previous policy of enforcement discretion. The intended implications for software, particularly AI-driven clinical decision support (CDS) tools that are integral to LDT workflows, were deep. Many such software components, previously considered part of the “practice of medicine” within a laboratory, would have fallen under the definition of a medical device, specifically a Software as a Medical Device (SaMD). This would have necessitated a shift in how these software components are designed, validated, and maintained. Companies would have needed to consider a complete Quality Management System (QMS) / ISO 13485 framework for their software from inception, ensuring compliance with GMLP (Good Machine Learning Practice) principles throughout the development lifecycle.
Dual Oversight Challenges for AI Platforms Like Tempus AI
Platforms like Tempus AI exemplify the dual oversight challenges that would have emerged from the FDA’s rescinded LDT rule. Tempus AI develops both LDTs and sophisticated clinical decision support tools that use genomic and clinical data to inform treatment decisions. Under the intended framework of the rescinded rule, not only would Tempus’s LDTs themselves have been subject to FDA premarket review, but the integrated AI software that processes, interprets, and delivers insights from these tests would also likely have been scrutinized as a SaMD. Consider a scenario where Tempus’s AI analyzes genomic data from an LDT to predict a patient’s response to a specific chemotherapy. Previously, the interpretation generated by the AI might have been considered part of the laboratory’s internal process. With the rescinded rule, the AI component, if it were to make a diagnostic determination or provide information essential for diagnosis, would have been classified as a regulated device. This would have meant:
- Premarket Review: The AI software would likely have required a 510(k) Clearance or, for novel applications, a De Novo Classification. This would have demanded strong clinical validation demonstrating the software’s analytical and clinical performance.
- Quality System Compliance: The development, validation, and maintenance of the AI would have needed to adhere to QSR, including rigorous documentation, risk management, and change control procedures.
- Post-Market Surveillance: The AI’s performance in real-world settings would have been subject to ongoing monitoring, with adverse event reporting requirements. The challenge would have been further compounded by the inherent adaptive nature of many AI/ML algorithms. Without a Predetermined Change Control Plan (PCCP), every time an AI model retrained on new data, a new premarket submission could have been required, leading to an unscalable regulatory burden. This highlights the critical need for AI-native companies to build regulatory compliance into their foundational architecture, rather than treating it as a post-development add-on.
Policy Coordination to Prevent Overlapping Requirements
The rescinded LDT rule, while intended to address a critical gap in diagnostic oversight, would have created a complex regulatory environment where policy coordination would have been paramount. The potential for overlapping requirements between laboratory oversight (CLIA, CAP) and medical device regulation (FDA) for the same integrated diagnostic solution was significant. For instance, a laboratory assay might already be subject to extensive validation under CLIA, while the AI software interpreting its results would have faced a separate, albeit related, FDA SaMD review under the rescinded rule. The Association for Molecular Pathology (AMP) had voiced concerns regarding parts of the LDT rule Association for Molecular Pathology public comment on LDT rule, emphasizing the potential for stifling innovation and increasing healthcare costs without commensurate patient benefit if regulatory burdens were to be duplicative or overly prescriptive. Regulators continue to carefully consider these concerns to ensure that any future framework encourages, rather than hinders, the responsible development and deployment of innovative AI health tools. Key areas requiring policy coordination include:
- Harmonization of Validation Standards: Exploring mechanisms to accept or simplify evidence generated for CLIA/CAP accreditation during FDA SaMD review, reducing redundant testing and documentation.
- Clear Jurisdictional Delineation: Providing unambiguous guidance on where the LDT rule ends and SaMD regulation begins, especially for integrated systems. This is particularly important for distinguishing between Clinical Decision Support (CDS) that is unregulated versus diagnostic AI that functions as a regulated device.
- Adaptive Regulation for AI/ML: Using existing frameworks like the PCCP to allow for predefined modifications to AI models without requiring entirely new premarket submissions for every iteration. This acknowledges the unique iterative development cycle of AI.
- Data Sharing and Real-World Evidence (RWE): Encouraging the use of RWE from real-world clinical deployment to inform both LDT and SaMD post-market surveillance and performance monitoring, fostering continuous improvement and safety. FDA guidance on Real-World Evidence for medical devices. While the FDA’s final rule to regulate LDTs was in the end vacated and rescinded, its issuance had highlighted the agency’s intent to redraw the boundaries for diagnostic software, potentially compelling companies to adopt a SaMD-informed architecture from the outset. For federal healthcare regulators and diagnostics policy experts, the imperative remains clear: proactive and thoughtful policy coordination is essential to navigate this evolving field, ensuring patient safety and promoting innovation without imposing undue and redundant regulatory burdens on the next generation of AI-driven health solutions. ** Methodology and Source Note: This analysis is derived from a deconstruction of the FDA’s proposed rule on Medical Devices. Laboratory Developed Tests as published in the Federal Register, and public comments submitted by industry groups such as the Association for Molecular Pathology. Data points regarding the timeline of the FDA LDT final rule and estimated number of tests affected are verified against official FDA publications and court documents.*
Frequently Asked Questions
What was the FDA’s rationale for the now-rescinded LDT rule?
The FDA’s rationale for the rescinded rule centered on the increasing complexity and risk profile of modern LDTs. Many of these tests now resemble commercially manufactured diagnostic kits. The agency sought to subject these LDTs, including those incorporating advanced computational analysis, to the same regulatory requirements as other medical devices.
How would the rescinded LDT rule have impacted AI-driven clinical decision support tools integrated into laboratory workflows?
Many such AI software components, previously considered part of the ‘practice of medicine,’ would have fallen under the definition of a medical device, specifically a Software as a Medical Device (SaMD). This would have necessitated a shift in how these software components are designed, validated, and maintained, requiring compliance with a comprehensive Quality Management System and GMLP principles.
What specific regulatory requirements would AI software, like that used by Tempus AI, have faced under the rescinded LDT rule?
AI software would likely have required premarket review (e.g., 510(k) Clearance or De Novo Classification), demanding robust clinical validation. Its development, validation, and maintenance would have needed to adhere to Quality System Regulations (QSR), and its real-world performance would have been subject to ongoing post-market surveillance and adverse event reporting requirements.
What was a significant regulatory challenge identified for adaptive AI/ML algorithms under the rescinded rule?
A significant challenge was the adaptive nature of many AI/ML algorithms. Without a Predetermined Change Control Plan (PCCP), every time an AI model retrained on new data, a new premarket submission could have been required. This would have led to an unscalable regulatory burden for companies.
What broader policy concern did the discussions around the rescinded LDT rule highlight for federal healthcare regulators?
The discussions highlighted the complex intersection of lab oversight and software validation, and the need for a coordinated policy response. Federal healthcare regulators continue to consider the need to prevent overlapping and potentially contradictory requirements between laboratory oversight (CLIA, CAP) and medical device regulation (FDA) for integrated diagnostic solutions.
