This site is intended for UK healthcare professionals
Medscape UK Univadis Logo
Medscape UK Univadis Logo

Urine Test Identifies Endocrine Hypertension With Help From AI

GLASGOW, Scotland – A machine-learning model can use the results of a simple urine test that measures steroid hormones to assess the likelihood of a person’s hypertension being linked to excess cortisol and aldosterone. 

Alessandro Prete, MD, a consultant endocrinologist from the Institute of Metabolism and Systems Research, University of Birmingham, presented these findings at this year’s meeting of the Society for Endocrinology (SfE BES) 2023.

“Here we have developed a test that only uses urine. It’s noninvasive but at the same time it is very accurate at detecting hypertension associated with Cushing’s syndrome and primary aldosteronism,” he told Medscape News UK. 

Prete stressed the need to find a better means of detection because, “Currently, less than 1% of people with primary aldosteronism are tested for hypertension. So imagine how many patients are not detected and therefore not treated properly.”

He explained that many people currently go undetected because the existing process is difficult to conduct in a standard clinical setting, requiring blood, urine, and saliva sampling. “Being able to offer a test that is easy to do and convenient for the patient would be ideal.” 

“Detecting this type of hypertension is important because we know how to treat it through surgical removal of the adrenal glands or via medication,” said Prete, adding that “without treatment, it can be very serious with adverse outcomes, including possible cardiovascular events as well as other morbidity and mortality”. 

Large Multinational Translational Research

Prete and his colleagues from across nine European countries and parts of Asia developed their test using resources across multiple disciplines, including mathematics, data science, computing, metabolomics, cardiology, and endocrinology.

They used mass spectrometry to profile multiple steroids in the blood and urine of 1400 hypertensive adults with and without endocrine causes. Twenty-seven steroid metabolites were analysed, including androgens and androgen precursors, mineralocorticoids, glucocorticoid precursors, and glucocorticoids, all produced by the adrenal glands. The aim was to look for signs of aldosterone or cortisol association with the patient’s hypertension. 

Machine learning simplified the analysis of the many thousands of measurements taken. 

The area under the curve (AUC) was used to indicate test accuracy: the closer the AUC is to 1 then the better the test. Multiple urine steroids from patients with hypertension caused by primary aldosteronism, Cushing’s syndrome, or pheochromocytoma, were compared to profiles from patients with essential hypertension. 

“For patients with Cushing’s-related hypertension, the AUC was 0.93, while for patients with primary aldosteronism, the AUC was 0.73, which increased to 0.86 when compared only to patients with low-renin essential hypertension. These results suggest that this is potentially a highly accurate test to diagnose Cushing’s syndrome and primary aldosteronism,” said Prete. 

He highlighted the convenience of the test for patients and clinicians because it only requires a urine sample and no blood tests. It could also be done at home and posted to a central laboratory for testing. 

The test results provide a risk score as to the likelihood of the patient having Cushing’s or aldosterone-related hypertension, or that they have standard hypertension, explained Prete. 

Going forward, the researchers want to train and refine the model on the leading steroid markers to aid translation into clinical practice. They would also like to incorporate other parameters into the model, including renin, potassium, adrenocorticotropic hormone (ACTH), as well as train the model to distinguish ACTH-dependent versus ACTH-independent Cushing’s. 

Promising Potential

Robert Semple, professor of translational molecular medicine at the University of Edinburgh, Scotland, and a practising endocrinologist who was not involved in the research, commented: “This study is exciting because of its scale, featuring both a training and a test set of samples, the application of machine learning to complex steroid profiles, and the finding that analysis of urine samples may discriminate both Cushing’s syndrome and primary aldosteronism well from primary hypertension, especially in conjunction with testing of blood renin concentration.”

“With further refinement, this test offers the real promise of early noninvasive diagnosis of conditions that may be undiagnosed or take a long time to manifest overtly,” he told Medscape UK. 

Cristina Ronchi, MD, consultant endocrinologist and researcher also based at the University of Birmingham, but not linked to the study, added: “I think the test has a lot of potential clinical application, and the study itself shows the importance of collaboration in study work because this could not have been achieved by a single centre or even in a single country. It also underlines the importance of interdisciplinary work, for example clinicians, including endocrinologists, working with computer scientists, biochemists, experts in mass spectrometry. It’s also a good example of translational research, showing bench to bed application,” she said. 

Prete reports research contracts with HRA Pharma, Recordati, and Diurnal, and consulting fees from Recordati and Diurnal. Ronchi and Semple report no relevant disclosures.