📄 medicine

Change of the monitor T wave per mmol/L of potassium within patients: an analysis in emergency department patients with potassium below 6 mmol/L

In a retrospective analysis of emergency department patients with potassium levels below 6 mmol/L, the study found no statistically significant or stable association between changes in serum potassium and monitor T-wave amplitude, as the estimated effect was small, uncertain, and heavily influenced by a few data points.

Original authors: Leo Probst

Published 2026-10-20

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Leo Probst

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the emergency department, patients with potential heart or kidney trouble are often connected to a continuous monitor that displays their heart rhythm on a screen. This screen shows a wave called the T wave, which represents the heart resetting itself after each beat. At the same time, a blood sample is drawn to measure the level of potassium, a vital mineral that helps control how the heart muscle works. A well-known fact in medicine is that potassium levels can change the shape of this T wave, but there is a frustrating gap in time: the blood test results often take an hour or more to arrive, while the monitor shows the heart's activity instantly. Doctors have long wondered if they could simply watch the T wave on the monitor to see if potassium levels were shifting before the lab results came back, but it has been unclear exactly how much the wave changes for every small shift in potassium, especially when levels are not dangerously high.

A researcher at Johannes Kepler University of Linz set out to answer this specific question by looking at real data from an emergency department in California. The study focused on a specific group of patients: those whose potassium levels were below a high threshold of 6.0 millimoles per liter, who had a regular heartbeat, and who were not receiving treatment specifically to change their potassium levels. The researcher paired blood samples taken from the same patient during a single visit, usually one to twelve hours apart, and compared the change in potassium between the two samples with the change in the height of the T wave on the monitor during that same time. To do this, a computer program automatically measured the height of the T wave on the monitor recordings, looking for a pattern where a rise or fall in potassium would correspond to a predictable rise or fall in the wave's height.

The analysis examined thousands of pairs of blood samples and monitor recordings, but only a small fraction met the strict criteria for the main question. In the end, the study looked at 216 pairs from about 200 patients. The results showed that for every increase of one millimole per liter in potassium, the height of the T wave on the monitor changed by an average of 0.26 millimeters. However, the researchers could not be certain that this small change was a real effect caused by potassium. The data was too scattered to rule out the possibility that the change was zero, or even slightly negative. The study found that the result depended heavily on just eleven specific patient pairs; when those few pairs were removed from the calculation, the estimated change dropped to nearly zero. This suggests that the apparent connection between potassium and the T wave height was not a stable, reliable pattern across the group.

The researchers also checked how consistent the computer's measurements were by having a human expert review some of the same recordings. Crucially, this expert was the study's author, who had also developed the computer algorithm and knew the hypothesis being tested. They found that the computer sometimes measured the T wave as smaller than the human did, particularly in the few pairs that had the biggest influence on the results. This difference in measurement, combined with the fact that the same person designed the tool and interpreted the data, added to the uncertainty. While the study did find a positive link between potassium and the steepness of the T wave, this finding was also unstable and changed depending on how the data was analyzed. Ultimately, the study concludes that in this group of patients with potassium levels below 6.0 millimoles per liter, there is no clear, reliable evidence that watching the T wave on a monitor can tell a doctor how much the potassium level has changed. The data is compatible with a very small change, but it is also compatible with no change at all.

This work does not prove that the T wave is useless, nor does it prove that it is a reliable guide. It simply shows that in patients who are not being treated for high potassium and whose levels are not extremely high, the relationship is too weak and too noisy to be used for individual decisions. The researchers note that their findings apply only to this specific group and do not extend to patients with irregular heartbeats or those with very high potassium levels who are receiving treatment. To get a clearer answer, future studies would need to look at patients with larger changes in potassium, such as those undergoing dialysis, and would need more precise ways to match the exact time of the blood draw with the exact moment the heart rhythm was recorded. For now, the monitor T wave remains an unreliable indicator of potassium changes in the emergency setting, and doctors must still wait for the laboratory results to know for sure.

It is important to note that this research has not yet been peer-reviewed by other independent scientists, as it is currently published as a preprint. Additionally, the study was not prospectively registered with a public database before it began; instead, the analysis plan was fixed and locked in a computer script only after the data was collected but before the main analysis was run.

Technical Summary: Change of the Monitor T Wave per mmol/L of Potassium Within Patients

Problem Statement
In emergency departments (ED), patients with potential potassium disorders are often placed on continuous lead II cardiac monitoring, yet serum potassium results are frequently delayed by an hour or more. While the physiological link between potassium and cardiac repolarization is well-established, the magnitude of T-wave amplitude changes within a single patient as potassium fluctuates within the near-normal range (specifically below 6 mmol/L) remains unclear. Existing literature suggests the ECG is an unreliable marker for potassium at a single time point and that T-wave changes are most evident during large shifts, such as those occurring in hemodialysis. This study aims to quantify the change in monitor T-wave amplitude per mmol/L of potassium change within individual ED patients who are not undergoing specific potassium-directed treatment.

Methodology
This is a retrospective, within-patient analysis of the MC-MED dataset (Stanford University, 2020–2022), comprising 118,385 adult ED visits. The study design and analysis script were fixed prior to the primary analysis run.

Data Selection: The analysis focused on consecutive serum potassium pairs from a single visit, separated by 1 to 12 hours. Primary inclusion criteria required an initial potassium level below 6.0 mmol/L, a predominantly regular rhythm, and no documented indication for potassium-directed treatment (though 34% of medication orders lacked administration times, so "no treatment indication" does not strictly equate to "untreated").
ECG Measurement: For each potassium sample, a 10-second lead II ECG segment nearest to the order time was selected. An automated algorithm (500 Hz sampling, using NeuroKit2) analyzed a 60-second window around this segment to generate a median beat. The algorithm measured T-wave amplitude (in mm, relative to the PR baseline and QRS deflection) and T-wave steepness (descending limb slope in mV/s). Valid windows required at least 20 seconds of good beats and <30% irregularity.
Statistical Analysis: The primary outcome was the slope ( $\beta$ ) of the change in T amplitude (second minus first) regressed on the change in potassium ( $\Delta K$ ), adjusted for heart rate changes using a spline. Estimates were derived via least squares with 2,000 patient-cluster bootstrap draws for 95% confidence intervals (CI).
Validation: A post-hoc "reader check" was performed by the author (who specified the algorithm) on 11 influential pairs (those whose removal significantly altered the slope) and 11 random comparison pairs. The reader measured T amplitudes blind to potassium values and algorithm outputs.

Key Results

Primary Cohort: Out of 5,585 usable pairs, 216 (3.9%) met the primary criteria, representing approximately 200 patients. The median initial potassium was 4.2 mmol/L, and the standard deviation of potassium change ( $\Delta K$ ) was 0.70 mmol/L.
T-Wave Amplitude Slope: The estimated change in T amplitude was 0.26 mm per mmol/L of potassium (95% CI: -0.04 to 0.63 mm).
- The result is considered imprecise; the interval includes zero, and the smallest slope detectable with 80% power was 0.40 mm per mmol/L.
- The estimate was highly dependent on 11 influential pairs. When these were removed, the slope dropped to 0.04 mm (95% CI: -0.10 to 0.23). Using a robust (Huber) estimator, the slope was 0.07 mm.
Measurement Discrepancy: In the post-hoc reader check, the automated algorithm measured significantly smaller T amplitudes than the human reader at the influential pairs (mean difference 0.13 mV), particularly where T waves were flat.
T-Wave Steepness: A secondary analysis of T-wave steepness showed a positive slope of 0.75 mV/s per mmol/L (95% CI: 0.21 to 1.37). However, this result was unstable across sensitivity analyses and lacked an independent reference reading.
Noise and Variability: The standard deviation of T-amplitude change unexplained by potassium was 1.4 mm, significantly higher than the within-sample repeatability limit (0.53 mm robust), indicating substantial variability between samples.

Significance and Claims
The paper concludes that in this specific, selected group of ED patients (initial potassium <6 mmol/L, regular rhythm, no treatment indication), the association between potassium changes and monitor T-wave amplitude changes is neither established nor excluded.

Modest Findings: The data are compatible with a mean change ranging from slightly negative to approximately 0.6 mm per mmol/L (with sensitivity analysis upper limits up to 1.0 mm).
Clinical Implication: The author explicitly states that no clinical recommendation follows from these results. The findings describe group-level statistics and do not provide a detection limit for individual patients or inform what a clinician might observe on a monitor in real-time.
Limitations: The study acknowledges that the primary pairs represent a small fraction of the total data, that sampling times were approximated by order times, and that the analysis was conducted on a single dataset without an independent replication cohort. The results do not apply to patients with potassium $\ge$ 6 mmol/L, treated hyperkalemia, irregular rhythms (e.g., atrial fibrillation), or 12-lead ECGs.

The author asserts that future studies requiring larger potassium changes, documented sampling times, positive controls, and independent reference readings are necessary to resolve the question of whether monitor T-wave amplitude is a reliable surrogate for potassium fluctuations in this context.

Technical Summary: Change of the Monitor T Wave per mmol/L of Potassium Within Patients

More like this