Why Correct QT for Heart Rate?
The measured QT interval varies with heart rate: when the interval between successive R waves is shorter, QT generally shortens as well. In a concentration-QTc exposure-response study, this creates a practical problem. A change in heart rate across ECGs can produce a change in measured QT even if the exposure-related effect of interest has not changed. A correction formula attempts to express QT on a common scale, usually the scale corresponding to an RR interval of one second.
The correction is not a direct measurement of an individual’s underlying repolarization independent of heart rate. It is a transformation based on an assumed QT–RR relationship. The choice of formula matters especially when ECGs span a broad range of heart rates or when the assumed relationship does not match the study data.
For example, a QT of 400 ms and an RR of 0.64 s give a QTc in milliseconds. It is not necessary to convert QT to seconds. By contrast, inserting RR as 640 milliseconds would produce an incorrect result; the RR value must first be converted to 0.64 seconds.
Deriving the Two Corrections
Both formulas can be understood as versions of a power-law correction. Suppose the QT–RR relationship is approximated by \(QT=kRR^\beta\), where \(k\) is the QT value predicted at \(RR=1\) second and \(\beta\) describes how strongly QT changes with RR. Since RR is entered numerically in seconds, the expression is understood on that scale.
If the exponent \(\beta\) were known and appropriate for the data, dividing QT by \(RR^\beta\) would remove the modeled RR dependence:
The corrected value is then the predicted QT at the reference RR of one second. Bazett uses the fixed exponent \(\beta=1/2\); Fridericia uses \(\beta=1/3\). These are not derived from a particular study’s ECGs. They are conventional choices, and each imposes its own rate-dependence assumption.
This result also predicts the direction of residual dependence. If \(\beta>\gamma\), corrected QT increases with RR; if \(\beta<\gamma\), it decreases with RR. Since heart rate in beats per minute is \(60/RR\) when RR is in seconds, a trend with RR has the opposite direction from a trend with heart rate.
Worked Example: Calculating Both Corrections
A hypothetical ECG has QT = 400 ms and RR = 0.64 s. Calculate Bazett and Fridericia QTc.
For Bazett, \(\sqrt{0.64}=0.8\), so:
For Fridericia, \(0.64^{1/3}\approx0.8618\), so:
Both results are in milliseconds because QT was entered in milliseconds. The substantial difference between 500 ms and 464.2 ms reflects the different exponents, not a difference in the measured ECG. This calculation alone does not establish which correction is better for a particular study.
How the Formulas Behave Across Heart Rates
Bazett divides by a larger power of RR than Fridericia does. This difference becomes consequential as RR moves away from one second. For a given measured QT, Bazett’s denominator changes more rapidly with RR. Whether that produces overcorrection or undercorrection depends on the actual QT–RR relationship in the population and setting; neither formula is guaranteed to remove rate dependence everywhere.
Worked Example: Comparing Corrections Under a Hypothetical QT–RR Relationship
Suppose, purely for illustration, that QT follows \(QT=400RR^{0.4}\), with QT in milliseconds and RR entered in seconds. This relationship has an exponent between the Bazett and Fridericia exponents. Compare the corrections at RR = 0.5 s and RR = 1.5 s.
At RR = 0.5 s, the measured QT predicted by the assumed relationship is:
The corrections are:
At RR = 1.5 s, the predicted measured QT is:
Thus:
The assumed exponent is 0.4. The Bazett residual exponent is \(0.4-0.5=-0.1\), so Bazett QTc decreases as RR increases. The Fridericia residual exponent is \(0.4-1/3\approx0.0667\), so Fridericia QTc increases as RR increases. Neither correction is perfectly independent of RR in this example, although Fridericia’s exponent is closer to the assumed relationship. This is a mathematical illustration, not a claim that the hypothetical exponent applies to a clinical study.
Checking QTc for Residual RR Dependence
A correction should be evaluated in the ECG data to which it will be applied. A useful first diagnostic is a plot of QTc against RR, with points identified by participant when repeated ECGs are available. A visible systematic trend suggests that the correction has not removed the association with heart rate. Summaries or exploratory regressions of QTc on RR can help quantify the direction and size of a trend, but a pooled trend can reflect differences between participants as well as within-participant changes. It should not automatically be interpreted as a within-participant relationship.
The following hypothetical records illustrate a diagnostic comparison. The same participant has four ECGs over a range of RR values. The records are shown to make the arithmetic and the trend transparent; four observations are not sufficient to establish a population-level correction or make a safety conclusion.
Worked Example: Diagnosing Residual Dependence in Hypothetical ECG Data
For each record below, calculate both corrected values. RR is in seconds and QT is in milliseconds.
| ECG | RR (s) | QT (ms) | Bazett QTc (ms) | Fridericia QTc (ms) |
|---|---|---|---|---|
| 1 | 0.64 | 343 | 428.8 | 398.0 |
| 2 | 0.81 | 378 | 420.0 | 405.5 |
| 3 | 1.00 | 410 | 410.0 | 410.0 |
| 4 | 1.21 | 443 | 402.7 | 415.7 |
For ECG 2, the Bazett calculation is \(378/\sqrt{0.81}=378/0.9=420.0\) ms. The Fridericia calculation is \(378/(0.81)^{1/3}\), where \((0.81)^{1/3}\approx0.9322\), giving \(405.5\) ms. For ECG 4, the corresponding calculations are \(443/\sqrt{1.21}=443/1.1=402.7\) ms and \(443/(1.21)^{1/3}\approx443/1.0656=415.7\) ms. ECGs 1 and 3 are calculated in the same way; at RR = 1 s both formulas leave QT unchanged.
Across these records, Bazett QTc falls from 428.8 to 402.7 ms as RR rises from 0.64 to 1.21 s. Fridericia QTc rises from 398.0 to 415.7 ms over the same interval. The opposing trends are consistent with an underlying QT–RR exponent between 1/3 and 1/2 in this illustrative series: the Bazett exponent is too large for this pattern, while the Fridericia exponent is too small. For a rough check, the endpoint exponent implied by the uncorrected QT values is:
This endpoint calculation is descriptive and sensitive to measurement error; it is not a reliable estimate of a study-wide exponent from four ECGs. In a real study, inspect the full distribution, participant patterns, measurement quality, and range of heart rates. If residual dependence is substantial, the correction may be inadequate for the intended analysis.
Assumptions, Alternatives, and Failure Modes
Using either fixed correction assumes that a single exponent is an adequate approximation over the observed RR range and that the ECG measurements are suitable for this transformation. A further practical assumption is that rate at the recorded ECG is relevant to the QT interval being corrected. QT adaptation to a changing heart rate may be delayed, so an ECG recorded soon after a rate change can reflect more than the contemporaneous RR. Irregular rhythms, poor-quality waveforms, and inaccurate interval measurements can also undermine interpretation. The formulas do not repair these problems.
A practical diagnostic sequence is:
Confirm the QT and RR measurements, use RR in seconds, and verify that QT and QTc are reported in the same unit.
Use the same ECG records and apply the formulas without mixing RR units or rounding intermediate values excessively.
Plot each corrected value against RR and, where repeated records exist, look at participant-level patterns as well as the pooled pattern.
Consider whether residual trends, the range of RR, and ECG timing make the selected correction reasonable for the planned exposure-response interpretation.
A common alternative is to model measured QT as a function of RR rather than imposing either fixed correction exponent. Such an approach can allow the data and model structure to represent rate dependence more directly, but it requires appropriate specification and adequate data over the RR range. It is not automatically superior: a flexible relationship can be imprecise with sparse data, and a model must still address the study’s sampling structure and measurement quality. For a fixed correction, Bazett and Fridericia are straightforward and widely recognizable, but their convenience does not establish their adequacy for every dataset.
Why the Choice Matters in Exposure-Response Studies
When concentration and heart rate vary across ECG records, residual RR dependence in QTc can complicate interpretation: an apparent pattern across records may partly reflect heart-rate differences rather than the exposure-response relationship of interest. Conversely, selecting a correction only because it gives a preferred result is not defensible. The correction should be chosen and evaluated with reference to the study data, the range of RR, and the planned analysis. Reporting the formula, the units, and the residual-dependence assessment makes the analysis more interpretable.
Neither formula converts QT into a universal measure free of physiologic context. A small apparent trend in a narrow RR range may be difficult to distinguish from ECG variability, while the same correction may perform poorly over a wide range. Results should therefore be interpreted as corrected QT under a stated convention, not as proof that all heart-rate effects have been removed.
Check Your Understanding
Answer each question, then compare with the solutions.
- A hypothetical ECG has QT = 360 ms and RR = 0.81 s. What are its Bazett and Fridericia QTc values?
- If the true QT–RR exponent were 0.45, would Bazett-corrected QT tend to rise or fall as RR increases? What about Fridericia-corrected QT?
- Why is it incorrect to enter RR = 800 when the measured RR is 800 ms?
- A plot shows QTc increasing with RR after Fridericia correction. What does that suggest about the correction in the observed data?
- Give one reason why a pooled QTc-versus-RR trend might not describe the within-participant relationship.
Solutions. (1) \(\sqrt{0.81}=0.9\), so Bazett QTc is \(360/0.9=400\) ms. Since \(0.81^{1/3}\approx0.9322\), Fridericia QTc is \(360/0.9322\approx386.2\) ms. (2) Bazett leaves exponent \(0.45-0.5=-0.05\), so corrected QT tends to fall as RR increases. Fridericia leaves exponent \(0.45-1/3\approx0.1167\), so corrected QT tends to rise as RR increases. (3) The formulas require RR in seconds; 800 ms must be converted to 0.8 s before substitution. (4) It indicates residual positive dependence on RR in those data; Fridericia has not fully removed the observed association, though measurement noise and study structure should also be examined. (5) Participants may differ in their average RR and QT levels, so a pooled association can combine between-participant differences with within-participant changes.