The Association for Diagnostics & Laboratory Medicine (ADLM) has called for updates to the Clinical Laboratory Improvement Amendments (CLIA) to address the growing use of artificial intelligence (AI) in laboratory medicine.
In comments submitted September 15 in response to a federal request for information on potential CLIA updates, ADLM recommended that AI-based tools remain subject to the same professional expertise, quality systems, validation requirements, and monitoring standards that apply to traditional laboratory testing.
CLIA has governed clinical laboratory testing in the United States since 1992. While traditional software has long been used to verify, interpret, and report laboratory results, ADLM noted that AI and machine learning introduce different potential failure modes that are not fully addressed by the current regulatory framework.
AI introduces new challenges for laboratory oversight
According to ADLM, conventional software errors generally affect patient cases that meet the same programmed conditions and can be evaluated using cases with known expected outputs. AI-based models, by contrast, may produce case-specific errors, making problems more difficult to identify and troubleshoot.
Generative AI presents additional concerns. ADLM noted that these systems can produce inaccurate or unsupported information, omit clinically important facts, or change their behavior following updates to a model, prompt, or knowledge base. Any AI-generated content incorporated into a laboratory report remains subject to the laboratory's quality and reporting requirements, according to the organization.
ADLM recommends that CLIA distinguish between conventional software and AI models and that laboratories account for those differences when validating and monitoring performance.
The comment letter also argues that validation requirements should reflect the technology being evaluated rather than apply the same approach to every testing method. “A mass spectrometry assay, NGS panel, and AI-based image analysis system cannot be evaluated using the same validation paradigm,” ADLM wrote.
For learned models, ADLM recommends evaluation using representative cases appropriate to the intended use. Depending on the type of output, performance assessment could include sensitivity, specificity, predictive values, agreement measures, calibration, and failure or abstention rates.
ADLM proposes a risk-based regulatory approach
The organization recommends that the Centers for Medicare & Medicaid Services and Centers for Disease Control and Prevention establish federal oversight for AI-based tools without unnecessary duplication. ADLM noted that the Food and Drug Administration may regulate individual software products or medical devices, while CLIA addresses the laboratory's responsibility for accurate and reliable testing.
ADLM also recommends a risk-based, technology-appropriate approach to AI oversight. Under this approach, CLIA would establish clear quality expectations while allowing laboratory directors and qualified laboratory professionals to determine scientifically appropriate methods for meeting those expectations.
“The most immediate need is greater clarity regarding how the existing CLIA framework applies to advanced software and AI rather than creation of a separate regulatory structure,” the comment letter stated.
The recommendations extend to facilities that independently analyze patient-specific laboratory data or provide specialized laboratory interpretation that generates or helps generate a clinical test result or interpretation. ADLM said these activities should be considered part of the total testing process and subject to appropriate CLIA oversight.
Laboratory oversight remains central
"AI has the potential to support tremendous advances in laboratory medicine, but innovation in this area must be balanced with the need to ensure test quality and patient safety,” ADLM President Stanley F. Lo, MD, said in a press release.
"Laboratories have the expertise and quality systems needed to evaluate, implement, and continuously monitor AI tools, making continued laboratory oversight essential to the responsible use of AI in clinical laboratory testing," he added.
Overall, ADLM recommends updating CLIA to provide greater clarity for advanced software and AI while preserving its flexible, performance-based framework. The organization said the goal should be to establish expectations proportionate to the technology and clinical risk rather than create a separate regulatory structure for AI.
This article was created with the assistance of Generative AI and has undergone editorial review before publishing.
