(Changes "he" to "she" in paras 12-14.)
By Marilynn Larkin
NEW YORK (Reuters Health) - Two ECG leads plus artificial intelligence may be all that's needed to detect hyperkalemia in at-risk patients with renal disease, researchers suggest.
"Over 30 million Americans with renal disease or heart failure are at risk for potentially life-threatening hyperkalemia due to their underlying diseases or the medications used to treat them," Dr. Paul Friedman of Mayo Clinic in Rochester, Minnesota, told Reuters Health by email.
"It has been long known that with marked hyperkalemia, ECG changes exist," he said. "What was not known was whether more subtle changes - before they were associated with potentially devastating elevations of blood potassium - could be detected. This led to a series of studies, initially in people on dialysis, that have to date demonstrated that the ECG can be effectively used to determine if a high blood potassium level is present."
"The US Food and Drug Administration approved this (approach) in the breakthrough device pathway, since it could enable what we've called a 'bloodless blood test,'" he noted. "People at home could determine whether hyperkalemia is present using their smart phones. If validated, patients could be alerted and then take a recently approved, safe potassium-lowering medication to prevent a potential catastrophe."
Dr. Friedman and colleagues trained a deep neural network using more than 1.3 million ECGs from about 450,000 patients seen at the Mayo Clinic in Minnesota from 1994 to 2017.
The network was trained using two or four ECG leads to detect serum potassium levels of 5.5mEq/L or less and was validated using retrospective data from Mayo Clinic cohorts in Minnesota, Florida, and Arizona. These cohorts included 61,965 patients with stage 3 or greater chronic kidney disease. Each patient had a serum potassium count drawn within four hours after their ECG was recorded.
About half (55%) of all ECGs were from men.
As reported online April 3 in JAMA Cardiology, the prevalence of hyperkalemia in the three validation data sets ranged from 2.6% in Minnesota to 4.8% in Florida. Using only ECG leads I and II, the model's area under the curve was 0.883 for Minnesota, 0.860 for Florida, and 0.853 for Arizona.
Using a 90% sensitivity operating point, the sensitivity and specificity, respectively, were 90.2% and 63.2% for Minnesota; 91.3% and 54.7% for Florida; and 88.9% and 55% for Arizona.
"Next steps will include testing the model in high-risk populations prospectively, and then have people use it themselves at home," Dr. Friedman said.
Dr. Michal Melamed, associate professor of medicine and of epidemiology and population health at Albert Einstein College of Medicine and Montefiore Health System in New York City, said the findings are "likely feasible and may be helpful in a few scenarios."
"A machine could be developed that patients could use at home to make sure their potassium levels are not too high," she said in an email to Reuters Health. "It could also be used in patients when they first arrive to an emergency room or in an ambulance, and if they do have high potassium levels, they may be able to be treated more quickly than waiting for labs to come back."
That said, she noted, "It is unclear if the ECG test can tell how high the potassium is, or just whether it is high. I would be concerned for patients to use this at home, as it can detect high potassium levels, but not low levels. If patients take a medication to lower potassium levels, it may go too low."
"This has a lot of potential as there are new medications for high potassium levels such as patiromer and zirconium cyclosilicate that have been recently approved by the FDA," she added. "More detection of high potassium levels will likely lead to more use of these medications. It is important to do trials that test whether treatment of high potassium levels at home improves outcomes."
The study was funded by AliveCor. Mayo Clinic licensed patent applications and technology know-how to AliveCor and invested in the company. Dr. Friedman and other Mayo coauthors may benefit financially from the technology's commercialization.
JAMA Cardiol 2019.