Listen to a discussion of this article through our AI-generated podcast.
Beginning with our September 2024 issue, CAP TODAY now offers the option to listen to a discussion of our articles with voices reminiscent of a Siri or Garmin speaker. Since this AI-generated podcast uses technology that is still in its early stages of development, there may be some errors in pronunciation and the tone may sound too upbeat for the material covered.
Leslie Williams
September 2024—Tried and true but also having untapped potential is how three industry insiders see urinalysis. Though traditional urinalysis serves its purpose well, they say, it’s easy to envision the next level.
“We tend to look at urinalysis with a fairly myopic view of counting particles and doing a dipstick,” says Jason Anderson, MPH, MT(ASCP), senior product manager for urinalysis solutions, Sysmex America. “So what becomes the definition of urinalysis in the future?” he asks. “When we look at blood, it isn’t plasma analysis, it isn’t serum analysis. It’s specific disease conditions, disciplines that use that fluid type. So I see urinalysis becoming a broader field in a sense that there has been a lack of research in urinalysis in general for some time and there’s huge capacity in a urine sample and an unmet need for better biomarkers.” This need, he says, spans the spectrum of diseases, not just renal diseases. “And those biomarkers, those metabolites, that are potentially useful in diagnosing various conditions can be found in urine.”
It comes down to research, Anderson says, “to understanding the pathophysiology better. How do these biomarkers work? When do they show up? Are they transient? Are they stable?” The questions are numerous, and as more emphasis is placed on urinalysis as a tool, “in a broad sense it will become more useful in diagnosing conditions going forward,” he says.
Researchers have identified the metabolite adenine as one such biomarker of progression in early diabetic kidney disease (Sharma K, et al. J Clin Invest. 2023;133[20]:e170341). “There are thousands of metabolites in urine, and as we start to understand those metabolites, we can look at concentrations,” Anderson says, noting there are therapeutics that can manage adenine production. “So if adenine contributes to the progression of diabetic kidney disease, then there’s reason to believe that if you could test for that in your urine and it’s at a certain level, the patient could then take the appropriate medication to reduce that level, which may in turn manage diabetic kidney disease or possibly prevent or reverse it.”
There are also known markers of metabolites in urine that are associated with Alzheimer’s disease, Anderson says. “So I see urinalysis certainly playing a role in personalized medicine, especially in light of the advancements in identifying proteomic and genomic biomarkers in urine, the development of molecular-based and other advanced-technology assays, and the increasing capabilities to tailor treatments to the metabolic profile. This isn’t anything I see in the immediate future, but certainly I see it at some point in the future.”
Today, multidrug-resistant pathogens and antibiotic resistance complicate treatment for infection. Ahmed Bentahar, MD, PhD, medical director, Beckman Coulter, says bacterial virulence factors could be used as biomarkers to flag infections in the early stage of infection when antibiotic therapy is most efficient. “These markers would be a huge advance for medicine,” Dr. Bentahar says, “because it would allow the development of targeted therapies that stop bacteria from evolving and progressing.” Identifying bacterial virulence factors may make it possible to develop vaccines and small-molecule therapies to target these factors, he says, decreasing pathogenicity and preventing chronic infection.
Also to come is the use of artificial intelligence, for which large-scale prospective studies are needed, say the authors of an article on AI’s applications in urinalysis (De Bruyne S, et al. Clin Chem. 2023;69[12]:1348–1360).
“Retrospective studies consistently demonstrate good performance of AI models in urinalysis,” they write, “showcasing their potential to revolutionize clinical practice. However, to comprehensively evaluate the real clinical value and efficacy of AI models, large-scale prospective studies are essential.”
Dr. De Bruyne and coauthors, of the Department of Laboratory Medicine at Ghent University in Belgium, examined the literature encompassing automated urine test strip and sediment analysis, UTI screening, and the interpretation of complex biochemical signatures in urine, including the use of mass spectrometry and molecular-based profiles. In all of it, they write, “AI represents a promising tool in urinalysis.”
Says Dr. Bentahar: “A thoroughly trained AI model can provide detailed insights about cell morphology, cell characterization, and functional characteristics of the cells, casts, and crystals in the urine. Ultimately, AI could improve the accuracy and efficiency of digital microscopy and may contribute to disease state detection, patient triage, disease severity assessment, patient outcome prediction, and treatment response monitoring.”
The “game changer,” he says, will be when AI can optimize urinalysis costs and be used for more precise microbial characterization to identify bacteria in urine “to the point where they can bypass or at least get a head start on culture and microbial sensitivity.”
Inbal Kinamon, MBA, global business line head for central urinalysis, Siemens Healthineers, agrees. “If urinalysis results indicate that a patient has a UTI and their medical history shows antibiotic resistance, AI could flag that and recommend alternative treatments. Alternatively, we have technology that allows us to subcategorize bacteria as rods or cocci, and AI could assist clinicians with their prospective prediction of what antibiotics to prescribe.”
Studies are identifying the connection between metabolic changes and chronic diseases like chronic kidney disease and Alzheimer’s disease, Kinamon says. “AI-enhanced digital microscopy could potentially help flag dysmorphic red blood cells in urine, changing the diagnostic pathway for chronic kidney disease and reducing unnecessary testing.” In theory, she says, AI-enhanced urinalysis could also provide early diagnosis of Alzheimer’s before symptoms appear, “but that will require gathering data from a very large population over an extended period of time.”
Combining urinalysis testing data with the electronic health record could generally improve predictive analytics, Kinamon says. “By integrating test results with patient demographics and medical history, we could train AI to be even better at predicting who’s at risk for a given disease.” This would require EHR and laboratory information systems vendors to collaborate and share information, which Kinamon views as a barrier. “We’d need such a high level of collaboration between different health care systems, as well as patient consent, so this degree of integration may be beyond our control.”
But the potential for AI-enhanced urinalysis to become an even more powerful clinical decision support tool is undeniable.
“I think in the coming years we will not move without AI,” Kinamon says. The three highest barriers to AI use in urinalysis, in her view? AI development brings greater clinical value, she says, “but it also increases the cost associated with that testing. Are payers going to reimburse you?” In addition, medical laboratory scientists have to trust the AI results, “and it takes time to build that trust.” And as with all laboratory procedures, safety and accuracy come first. “We have to be very cautious about that.”
“We are in a highly regulated environment. We need to be very careful that by adopting AI into laboratory processes we don’t sacrifice accuracy,” she says. “We also hope that CMS will create predictable and consistent reimbursement for FDA-authorized AI technologies that provide quantitative or qualitative data to the physician that is not otherwise available.” This technology provides an additional service to the patient, she notes, by helping the physician identify the best diagnosis and treatment.
Kinamon points out that while the FDA has yet to release a final guidance for AI and machine learning for medical devices, it has authorized nearly 900 AI/machine learning products as of May this year. “The current FDA regulatory framework ensures rigorous premarket review and postmarket surveillance for AI/ML-enabled medical devices, striking the right balance between incentivizing innovation and protecting patient safety,” she says.
Meanwhile, the FDA Center for Biologics Evaluation and Research is participating in public workshops and discussions to better understand the clinical uses of AI/ML and to inform an appropriate regulatory framework.
The FDA’s decision on the use of AI in the clinical lab, Dr. Bentahar says, “depends on whether we are claiming that AI will be able to make a diagnosis or we’re claiming that AI is just a support tool,” an adjunct to assist with interpreting results or integrating results from different disciplines. “As long as AI is presented as an adjunct to all the other tools the clinician or laboratorian is using and has the required accuracy and precision,” he says, “I don’t think it’s a big challenge for the FDA to approve its use.”
For any discipline, there are different levels of expertise, Dr. Bentahar says. And an integrated diagnostic approach that combines disciplines and expertise “and can be improved and facilitated by AI” holds promise for standardizing practice.
“AI can make results more quickly available to the clinician—who will ultimately make the differential or final diagnosis—to help establish a therapeutic plan.”
Being able to analyze vast patient data sets “will undoubtedly uncover unexpected findings,” Anderson says. Using the information responsibly, understanding its correlation to conditions and outcomes, and addressing the cost implications for early disease detection will be important work, he says, on the other side of which lies greater clinical value.
Dr. Bentahar is of the view “there are no boundaries for future technological advances and capability for urinalysis applications installed in tools used daily,” to provide more accurate diagnostics that shorten turnaround times and the response to prevent disease progression.
Leslie Williams is a writer in Temple, NH.