JACC: CLINICAL ELECTROPHYSIOLOGY VOL. -, NO.

-, 2017
ª 2017 BY THE AMERICAN COLLEGE OF CARDIOLOGY FOUNDATION ISSN 2405-500X/$36.00

PUBLISHED BY ELSEVIER http://dx.doi.org/10.1016/j.jacep.2016.12.005

A New QT Interval Correction Formulae to

Adjust for Increases in Heart Rate
Simon W. Rabkin, MD,a Elena Szefer, MSC,b Darby J.S. Thompson, PHDb,c

ABSTRACT

OBJECTIVES This study sought to develop a formula from a large population-based study that best fit the QT-heart
rate (HR) relationship independent of the standard mathematical relationships.

BACKGROUND Attempts to adjust or correct for the impact of HR on the QT interval (QTc) have applied various
mathematical equations to electrocardiographic (ECG) data rather than allowing the data to determine the form of the
relationship.

METHODS A spline correction function was developed using the ECG data from NHANES (National Health and Nutrition
Examination Surveys) II and III. The magnitude of linear, quadratic, and cubic trends in the relationship between HR and
each QTc was quantified using an F-statistic with differences between QTcs compared using a permutation procedure.

RESULTS The effect of HR on QT was obliterated by the spline QT for both men and women. The cross-validated spline
QTc was superior (i.e., flatter) to 6 other formulae, including ones proposed previously. In ECGs from the clinic with HRs
faster than 70 beats/min, the QTcs from different formulae were significantly (p < 0.0001) different from one another.
Individual values suggest the use of the longest and shortest QTc intervals as developed originally. The new QTc and its
population percentile ranking can be provided for clinical ECGs.

CONCLUSIONS A new QTc formula was developed which eliminates the relationship between QT and HR. At faster
HRs, the 2 most commonly used QTcs provide numerical values at the extremes of QTc. Compared to existing formulae,
the new formula had the best performance. (J Am Coll Cardiol EP 2017;-:-–-) © 2017 by the American College of
Cardiology Foundation.

A ssessment of the QT interval has attracted

considerable attention as a marker for poten-
tial drug-induced cardiac
drug-induced sudden cardiac death (1), electrolyte ab-
toxicity or
root of the heart rate, a concept which was proposed
earlier in studies by others on the duration of mechan-
ical systole with different heart rates. In the same
year, Fridericia (5) published his conclusion, on the
normalities (2), and identification of individuals with basis of 50 apparently normal individuals, that QT in-
genetic mutations that carry a high risk of sudden terval was a function of the cubed root of the RR inter-
death (3). A critical aspect of the evaluation of the val. Subsequently, many other QT adjustment
QT interval is its adjustment for heart rate because formulae have been proposed (6). The Framingham
of the recognized inverse relationship between QT in- baseline ECG data were used to develop a linear
terval and heart rate. The formulae most frequently regression model (7). A power function QT correction
used to produce a QT interval that corrects (QTc) for formula was derived from ECGs that were baseline re-
the changes in heart rate were developed almost 100 cordings, before drug treatment in Eli Lilly clinical tri-
years ago by Bazett (4) (QTcBZT) and Fridericia (5) als by Dmitrienko et al. (8) (QTcDMT). Most recently,
(QTcFRD). Bazett (4), in 1920, concluded that the rela- Rautaharju et al. (9) developed 2 formulae (QTcRTHa
tionship between the duration of systole shown on the and QTcRTHb) on the basis of data pooled from popu-
electrocardiograph (ECG) was a function of the square lation surveys as well as ECGs from baseline before

From the aDepartment of Medicine, Division of Cardiology, University of British Columbia, Vancouver, British Columbia, Canada;
b
Emmes Canada, Burnaby, British Columbia, Canada; and the cFaculty of Health Sciences, Simon Fraser University, Vancouver,
British Columbia, Canada. The authors have reported that they have no relationships relevant to the contents of this paper to
disclose.

Manuscript received July 5, 2016; revised manuscript received December 3, 2016, accepted December 15, 2016.
2 Rabkin et al. JACC: CLINICAL ELECTROPHYSIOLOGY VOL. -, NO. -, 2017
New QTc formula - 2017:-–-

ABBREVIATIONS drug testing. These studies used linear or measurements were made by a computerized ECG
AND ACRONYMS power-based equations and found the best analysis algorithm which eliminated intraobserver
parameters from the chosen equation. The variability (10,11).
NHANES = U.S. National
Health and Nutrition
mathematical equation, linear, exponential,
DEVELOPMENT OF A NEW QTc FORMULA. The new
Examination Survey cubic, and so forth, have been applied to
QT correction function, the spline QT correction, was
QTc = QT corrections for ECG data rather than allowing the data to
developed using R statistical software (2014, R Core
heart rate determine the form of the relationship. The
Team, Vienna, Austria). The spline correction function
QTcBZT = QTc proposed by first objective of this study was to develop a
Bazett was modeled using a cubic regression spline with 4
formula from a large population based study
QTcFRD = proposed by knots and a fixed adjustment for sex, with each
which best fit the QT-heart rate relationship
Fridericia observation weighted by the respective NHANES
independent of the standard mathematical
QTcDMT = proposed by sampling weight with spline parameters selected as
relationships, which presuppose a linear,
Dmitrienko et al. those that minimized the squared vertical distance to
power, or logarithmic function. The second
QTcFRM = QTc based on the fitted line (i.e., least squares estimate). The spline
Framingham data objective was to compare the developed for-
QTc was then computed by taking the difference be-
QtcHDG = proposed by Hodges
mula to previously derived QTc formulae.
tween QT measurements and the predicted value of
et al. The third objective was to define age- and
the regression spline at the observed heart rate and sex
QTcRTH = proposed by sex-based mean
SD for the formulae; the
and then adding the predicted spline value at a heart
Rautaharju et al. fourth objective was to test the new formula
rate of 60 beats/min for men. The predicted spline
using ECGs from a clinical setting.
value at a heart rate of 60 beats/min for men was used
as the “baseline,” where the spline-predicted value
METHODS
was considered to be deviation from this baseline.

POPULATION SAMPLE. The U.S. NHANES (National

An age-adjusted spline QTc was computed simi-

Health and Nutrition Examination Survey) II and III larly by adding a fixed adjustment for age to the

studies, conducted by the U.S. Centers for Disease spline correction function regression model. To

Control and Prevention (CDC), were chosen to eval- compute the age-adjusted QTc, the difference be-

uate the QT interval in different age groups. These tween QT measurement and predicted value of the

studies were selected because each NHANES survey regression spline at the observed heart rate, sex, and

was conducted using a representative sample of the age was computed, and then the predicted spline

civilian U.S. population, and importantly, weighting value at a heart rate of 60 beats/min for men at 50.3

factors were available, meaning that the results are years of age was added. The age of 50.3 years was

more representative than voluntary survey samples chosen as the baseline because it was the weighted

or volunteers in a drug trial. NHANES II and III 2012 estimate of mean age in the NHANES data.

and 2015 data were downloaded from the CDC (10,11) Although we were able to compute an age-adjusted
and from U.S. Department of Housing and Urban spline QTc formula, existing QTc formulae do not
Development 2008 data. Data were imported and account for age, and for fair comparison purposes, we
processed using Excel software (2013, Microsoft, compared the performance of the spline QTc, which
Redmond, Washington). The possibility that the same accounts for sex but not age, to the existing QTc
individuals were selected for both studies was formulae. Some results from the age-adjusted spline
considered not significant because of the randomness QTc are included to illustrate the benefit of addi-
of the participant selection process. Subject sampling tionally adjusting for age.
weights for NHANES I and II provided by the CDC An applet to compute the spline QTc from user
were used in all analyses. Data from NHANES II and input was made using the shiny package, a software
III were pooled to create one large data set. Then, package which builds web applets for data analysis
exclusion criteria were applied to exclude ECG fac- and computation using user-inputted data, in R
tors, known to affect the duration of the QT interval. software (12). Percentiles of the input data were
Briefly, the exclusion criteria removed subjects with: computed by comparing their QTc to the QTcs of the
1) no valid QT interval duration or heart rate data; 2) NHANES subjects of the same sex as the reference;
probable or possible myocardial infarction or major those who had higher QTc percentiles are indicative
ECG abnormalities; 3) rhythm not being in sinus; 4) of greater QT prolongation than usual.
probable left ventricular hypertrophy; and 5) left or To evaluate the spline-based QT conversion func-
right bundle branch block. The full definitions of all tion, we used a 10-fold cross-validation procedure to
ECG abnormalities are available in the NHANES obtain the spline QTc for individual subjects not
documentation (10,11). The heart rate and QT interval included in the training set for the spline QTc. In this
JACC: CLINICAL ELECTROPHYSIOLOGY VOL. -, NO. -, 2017 Rabkin et al. 3
- 2017:-–- New QTc formula

approach, spline QTcs were computed by randomly heart rates. To compare the performance of the spline
splitting the data into 10 folds of roughly equal size, QTc to other QTc formulae, we quantified the extent
called “test sets.” For each test set, all observations in to which trends remained between heart rate and QTc
the data that were not selected to be in the test set by using an F-statistic from separate, weighted
were included in a corresponding training set (i.e., 10 regressions on each QTc formulae and sex involving
training sets and 10 test sets). The spline correction linear, quadratic, and cubic functions of heart rate as
function was fit to each training set, as described predictors (14). The magnitude of the statistic is then
earlier, and cross-validated spline-QTcs were indicative of the strength of linear, quadratic, and
computed for the test set by using the training set fit. cubic trends in the relationship between heart rate
Because some data were held out of the training set in and QTc. If the correction formula succeeds in
each fold, the cross-validated spline QTcs were completely removing these trends, F will be 0. Large
computed using data that were not used to fit the values of F indicate that at least one of these re-
correction function and hence approximate perfor- lationships (linear, quadratic, or cubic) exists and that
mance in an independent data set. the QTc is not flat for all heart rates. To quantify dif-
ferences in F-statistic, a permutation testing proced-
COMPARISON OF DIFFERENT QTc FORMULAE. Six
ure was used to test whether the F-statistic from each
different heart rate correction formulae were evalu-
QTc formula was significantly higher than that of the
ated either because of their long-term usage or their
cross-validated spline QTc (15) in the N ¼ 13,627 ECGs.
recent introduction from large population studies.
Because some of the ECGs in the data have very
The formulae include the equations proposed by
extreme heart rates, a similar trimmed analysis was
Bazett (4) (QTcBZT), Fridericia (5) (QTcFRD), Hodges
conducted on the n ¼ 13,610 ECGs obtained at heart
et al. (13) (QTcHDG), Framingham (Sagie et al. [7])
rates between 40 and 120 beats/min.
(QTcFRM), Dmitrienko et al. (8) (QTcDMT), and Rau-
taharju et al. (9) (QTcRTHa), which were identified by
a proposed standardized nomenclature (6). Two RESULTS
formulae were described in Rautaharju et al. (9).
The relationship between QT interval and heart rate
CLINIC-BASED ECGs. Consecutive resting ECGs from demonstrated a consistent reduction in QT interval
a clinical reading session that fulfilled the criteria of with each incremental increase in heart rate. This
sinus rhythm and without bundle branch block, left inverse relationship between the duration of the QT
ventricular hypertrophy, ST-segment elevation interval and heart rate was evident in men (Figure 1A)
myocardial infarction, or significant ST-T wave and women (Figure 1B). The extremes of heart rate
changes were considered. The study was approved by were associated with a difference of at least 200 ms in
our Institutional Research Ethics Board. There were QT interval. The QT interval was longer at each heart
44 ECGs (50% men) that were obtained anonymously. rate in women than in men. The spline QT correction
Because most QTc formulae provide similar QTc formulae (see Online Appendix Equations) obliter-
values around a heart rate of 60 beats/min, ECGs with ated the effect of heart rate on QT in both men and
heart rate of 70 beats/min or greater were selected. women (Figure 2), where the regression lines up to a
No clinical information was available, similar to the cubic trend between QTc and HR are almost flat.
usual clinical ECG interpretation setting. ECGs were
acquired and digitally analyzed. Electrocardiographic COMPARISON OF SPLINE QTc WITH OTHER QT
waveforms were sampled at approximately 500 sam- CORRECTION FORMULAE. The cross-validated spline
ples per second, using the Marquette 12SL analysis correction function was compared with 6 different
program (GE Healthcare, Milwaukee, Wisconsin). The heart rate correction formulae QTcBZT, QTcFRD,
QT interval was measured “from the earliest detec- QTcHDG, QTcFRM, QTcDMT, and QTcRTHa. A superior
tion of depolarization in any lead (QRS onset) to the QT correction function is one in which the QTc has no
latest detection of repolarization in any lead (T offset) relationship to heart rate; that is, the mean QTc will be
(Marquette 12SL ECG Analysis Program; GE Health- the same or flat for all heart rates. Boxplots by heart
care). The QT interval and heart rate measured by the rate for each QTc were ordered by the p value
analysis program was used in the heart rate adjust- comparing flatness of the slope for each QTc formula
ment formulae. compared to the cross-validated spline formula for
On the basis of the requirement to adjust the QT men (Figure 3) and women (Figure 4). The fitted
interval for heart rate, a superior QT correction func- regression lines used to calculate the F-statistic for
tion is a QTc that has no relationship to heart rate; that each conversion formula that demonstrated the QTc-
is, mean QTc will be the same (graphically flat) for all heart rate relationship are displayed as the black and
4 Rabkin et al. JACC: CLINICAL ELECTROPHYSIOLOGY VOL. -, NO. -, 2017
New QTc formula - 2017:-–-

F I G U R E 1 QT Versus Heart Rate for Men and Women

Boxplots of the QT interval measurements for the NHANES cohort by heart rate, for men (left) and women (right). Boxplots are grouped into heart-rate intervals of
5 beats/min. The black horizontal line in the middle of each box denotes the median for the corresponding heart rate interval. The top and bottom of the box denote the
first and third quartiles for the heart rate interval, and the endpoints of the whiskers; the lines extending from the boxes denote 1.5 times the interquartile range from
the first and third quartiles. Outliers are points that fall beyond the whiskers and are plotted as points (the same approach is used in Figures 2 to 4). The QT-heart rate
relationship is shown by a solid line layered over the boxplot, which was constructed by fitted splines for each sex.

dashed lines. The regression lines displayed over the validated spline QTc succeeded for both men and
spline QTc and cross-validated spline QTcs are very woman at removing the linear, quadratic, and cubic
flat, demonstrating the lack of relationship between trends relative to most of the other formulae. The
HR and the spline QTc. trimmed analysis found more convincing results of
The permutation test indicates that in women the superiority of the cross-validated spline QTc than for
cross-validated spline QTc was superior (i.e., is the QTcFRD (ptrim ¼ 0.0041) and QTcFRM
flatter) to the sex-specific QTcRTH (p ¼ 0.045), (p trim ¼ 0.047) formulae.
QTCBZT, QTcDMT, QTcFRD, QTcFRM, and QTcHDG We next sought to examine the effect of age on the
formulae (p < 0.0001). There was no evidence new QTc formula because of the significant associa-
that the cross-validated spline QTc was superior to tion of QTc with age (16). There was evidence of linear
the general Rautaharju formula (p ¼ 0.49). The associations between the cross-validated spline QTc
trimmed analysis found similar results, that is, there and age in men and women (Figure 5). Weighted
was no evidence that the cross-validated spline linear models were fit with spline QTc as the
QTc was superior to the sex-specific QTcRTH outcome, and age was considered as a continuous
(ptrim ¼ 0.18). variable and the predictor. The spline QTc was noted
In men, the cross-validated spline QTc was supe- to increase by 0.21 (SEM ¼ 0.02) for a 1-year increase
rior to the QTcFRD (p ¼ 0.042), QTcBZT, QTcDMT, in age for men and by 0.16 (SEM ¼ 0.02) for a 1-year
QTcHDG, and QTcRTH formulae (p < 0.0001). There increase in age for women, a significant association
was no evidence that the cross-validated spline QTc for both men and women (p < 0.0001).
was superior to the sex-specific QTcRTH (p ¼ 0.49) or The methodology of using regression splines to
the QTcFRM (p ¼ 0.13) formulae in men. The cross- remove linear relationships with QT (such as the
JACC: CLINICAL ELECTROPHYSIOLOGY VOL. -, NO. -, 2017 Rabkin et al. 5
- 2017:-–- New QTc formula

F I G U R E 2 QT Interval Corrected for Heart Rate Using Regression Spline With Interquartile Ranges

Boxplots are for men (left) and women (right), grouped into heart rate intervals of 5 beats/min. Regression lines, with up to a cubic trend, between HR and
QTc are overlaid.

relationship between sex and QT) and nonlinear APPLICATION OF NEWER QT CORRECTION
relationships (such as the relationship between heart FORMULAE IN THE CLINIC. The values for the heart
rate and QT) can be applied to remove effects of other rate-corrected QT interval and their percentile
variables on the QT, such as age. By including a linear ranking can vary. As an illustration, the corrected QT
trend of age in the regression spline model of QT on interval for a 72-year-old woman with a heart rate of
heart rate and sex, an age-adjusted spline QTc can be 114 beats/min, the QTc varied widely. QTcBZT was the
computed. If a trend for age is included, the resulting longest at 469 ms, which was in the 92nd percentile,
age-adjusted spline QTc is no longer associated with whereas QTcFRD was 421 ms in the 23rd percentile. In
age (p ¼ 0.15 in a test for men; p ¼ 0.25 in a test for contrast, the spline QTc was 423 ms and was in the
women). The age-adjusted spline QTc performs simi- 63rd percentile (Table 1). We next examined ECGs
larly well in removing the relationship between heart from men with a heart rate over 70 beats/min. The
rate and QTc compared to other formulae, on the basis group had a mean
SD heart rate of 94.5
13.0 beats/
of the permutation test (p < 0.0001 compared to min and a QT of 358.1
23.5 ms. There was evidence
QTcBZT, QTcDMT, QTcHDG, and QTcRTHa in men and that the formulae were not equal (p < 0.0001;
QTcBZT, QTcDMT, QTcFRD, QTcFRM, and QTcHDG in Kruskal-Wallis test ¼ 52.9) in the calculated QTc
women; p ¼ 0.02 compared to QTcRTHb in women; p ¼ (Figure 6). The individual values suggest that the
0.002 compared to QTcFRD in men; p ¼ 0.25 compared longest QTc was shown by QTcBZT, whereas the
to QTcRTHa in women; p ¼ 0.02 compared to QTcRTHb shortest QTC was QTcFRD. Electrocardiographs from
in women; and p ¼ 0.44 compared to QTcRTHb in women with a heart rate over 70 beats/min were
men). The age-adjusted spline QTc, however, has an examined. The group had a mean
SD heart rate of
unfair advantage compared to the comparator 89.5
9.9 beats/min and a QT interval of 381.5
30.4
formulae because it additionally removes the effect of ms. The formulae were not equal (p < 0.0001;
age. For fair comparison, results and visualizations for Kruskal-Wallis test ¼ 34.6) in the calculated QTc
the spline QTc, which only adjusts for heart rate and (Figure 6). The individual values suggest that the
sex, are presented. longest QTc interval was shown by QTcBZT, thus the
6 Rabkin et al. JACC: CLINICAL ELECTROPHYSIOLOGY VOL. -, NO. -, 2017
New QTc formula - 2017:-–-

F I G U R E 3 For Men the QT Interval Corrected for Heart Rate for Different QTc Formulae

Boxplots of the QTc for each formula are plotted, grouped by heart rate in intervals of 5 beats/min. Both the spline QTc and the spline QTc from cross-validation (labeled
Spline QTc-CV) are included. The solid line is the fitted regression line from the regression of a linear, quadratic, and cubic trend on QTc for each QTc formulae. The
dashed line shows the fitted regression line on the subset of data with heart rates between 40 and 120 beats/min. Plots are annotated with the permutation p value,
which compares the F statistic for linear, quadratic, and cubic trends in each comparator formula with the F statistics for the cross-validated spline. The permutation
p value from the trimmed analysis, where only data from ECGs with recorded heart rates between 40 and 120 beats/min were used, are also included.
ECG ¼ electrocardiography.

most commonly used QTc formulae provide numbers wide range of different heart rates had the QTc
at the extreme of QTc measurement. calculated with each of the formulae (Figure 7). Not
In order to illustrate the effect of heart rate on only were there wide differences in QTc among the
various QTc formulae compared to the newly pro- different formulae at the heart rate of 80 beats/min
posed one, an individual with several ECGs over a but the discrepancies among the QTc formulae
JACC: CLINICAL ELECTROPHYSIOLOGY VOL. -, NO. -, 2017 Rabkin et al. 7
- 2017:-–- New QTc formula

F I G U R E 4 For Women the QT Interval Corrected for Heart Rate for Different QTc Formulae

Boxplots of the QTc for each formula are plotted, grouped by heart rate in intervals of 5 beats/min. Both the spline QTc and the spline QTc from cross-validation
(labeled Spline QTc-CV) are included. The solid line is the fitted regression line from the regression of a linear, quadratic, and cubic trend on QTc for each QTc formulae.
The dashed line shows the fitted regression line on the subset of data with heart rates between 40 and 120 beats/min. Plots are annotated with the permutation p value,
which compares the F statistic for linear, quadratic, and cubic trends in each comparator formula with the F statistics for the cross-validated spline. The permutation
p value from the trimmed analysis, where only data from ECGs with recorded heart rates between 40 and 120 beats/min were used, are also included as ptrim.
Abbreviation as in Figure 3.

increased at faster heart rates. For the spline QTc, the heart rate based on functionally agnostic modeling of
slope of the relationship between QTc and heart rate population ECG data with flexible regression splines.
was the smallest of all QTc formulae, and there were Previous formulae have used standard mathematical
no significant (p ¼ 0.84) differences from zero. In functional forms including exponentials, linear,
contrast, there were significant positive slopes in this and logarithmic functions and applied them to the
relationship for QTcBZT and QTcDMT as well as sig- QT-heart rate relationship (6). “Forcing the data”
nificant negative slopes for QTcFRM and QTcRTH. to fit a known relationship may have led to the
problem that existing QTc formulae do not accurately
DISCUSSION adjust for the impact of heart rate on the QT interval.
All QTc formulae are good at heart rates close to 60
This study presents a new formula for correction of beats/min, but very few formulae are also good at the
the inverse relationship between QT interval and higher and lower heart rates. The newly proposed
8 Rabkin et al. JACC: CLINICAL ELECTROPHYSIOLOGY VOL. -, NO. -, 2017
New QTc formula - 2017:-–-

F I G U R E 5 Comparison of the Mean Spline QTc by Age for Men and Women

Comparison of the mean spline QTc, with 95% confidence intervals, by age for men (left) and women (right). The linear regression line of age on spline QTc is plotted
with the 95% confidence interval.

formula did not presuppose a specific functional form to minimize or “correct” the effect of heart rate on the
(quadratic, logarithmic, or other) but rather used the QT interval began in 1920 with 2 simple mathematical
data to construct a new formula which appears to approaches, one dividing QT interval by the square
eliminate the impact of heart rate on QT interval. root (4) and the other dividing the QT interval by the
The new formula for the adjustment of the QT in- cube root of the heart rate (5). There was little if any
terval for heart rate was developed on the basis of electrophysiological basis for these approaches. Sub-
population data and a flexible regression spline sequent efforts to adjust the QT interval for heart rate
approach, permitting modeling of almost any shape of also relied heavily on mathematical relationships (6).
the QT-heart rate relationship. The history of attempts In addition, a nomogram has been developed using a

T A B L E 1 QT Interval and Percentile for Different QT Heart Rate Correction Formulae for a 72-Year-Old Woman With a Heart Rate
of 114 Beats/Min

Heart Rate, Heart Rate, Range,

Sex beats/min beats/min Age, yrs Age Group, yrs QT, ms Formula (Ref. #) QTc Percentile

Female 114 110–114 72 70–74 340 Spline 423 63

Female 114 110–114 72 70–74 340 Spline, age-adjusted 419 53
Female 114 110–114 72 70–74 340 Bazett (4) 469 92
Female 114 110–114 72 70–74 340 Dmitrienko et al. (8) 443 72
Female 114 110–114 72 70–74 340 Fridericia (5) 421 23
Female 114 110–114 72 70–74 340 Framingham (Sagie et al. [7]) 413 40
Female 114 110–114 72 70–74 340 Hodges et al. (13) 434 73
Female 114 110–114 72 70–74 340 Rautaharju et al. (9) 442 77
Female 114 110–114 72 70–74 340 Rautaharju, Gender (9) 431 54
JACC: CLINICAL ELECTROPHYSIOLOGY VOL. -, NO. -, 2017 Rabkin et al. 9
- 2017:-–- New QTc formula

F I G U R E 6 QT Values in a Data Set by Heart Rate in Men and Women

QT values in a data set of men (N ¼ 22) (upper panel) with mean
SD heart rate of 94.5
13.0 beats/min and a QT of 358.1
23.5 ms and in
women (N ¼ 22) with heart rate of 89.5
9.9 beats/min and a QT of 381.5
30.4 ms (lower panel). The individual QTc values (small points)
and their mean values (large point to the right of each cluster) are shown for each correction function.

heart rate of 60 beats/min as the reference value (17). some data were held out of the folds, the spline QTc
More recent formulae applied linear regression formula was cross-validated by applying it to data
modeling to obtain the parameters of a “best” fit linear that was not used to fit the correction function.
relationship between QT and heart rate (7,9). A linear, The new formula was compared to a number of
logarithmic or an exponential best fit of the data will different formulae, including both older and more
only be valid when the nature of the relationship is recently proposed formulae. The new formula was
truly known, and the relationship is indeed either a found to be an excellent approach to address the
linear, logarithmic, or exponential one. This is not the issue of the impact of heart rate on the QT interval.
case with the QT-heart rate relationship. Furthermore, We have previously used the approach that the slope
the relationship may change at different heart rates, of the relationship between heart rate and QT interval
for example, it may be linear in one heart rate range should be zero when there is no relationship between
and exponential in another range. A better way to fit QT and heart rate (20). Unfortunately, it is possible
such an unknown or potentially variable relationship for a zero slope of the QT-heart rate relationship to
is to use more flexible functions such as the spline arise if there is a systematic pattern in which positive
methodology (18). and negative values exist around the zero sloped QTc-
To address the concern that equations developed heart rate regression line (e.g., a sinusoidal pattern or
from epidemiologic studies are retrospectively accu- a “wobble” around a slope of zero). To quantify this
mulated and should have prospective evaluation (19), potential pathology, our current approach used the
we cross-validated the spline QTc formula. We used a F-statistic from a test of linear, quadratic, and cubic
novel approach in the field of QTc development, relationships and quantified both linear trends and/or
specifically, we developed the formula from a data set sinusoidal patterns and enabled comparison of “flat-
derived after randomly splitting the population ECGs ness” among QTc formulae. Statistical testing indi-
into 10 folds of approximately equal size. Because cated that there are fewer linear, quadratic, or cubic
10 Rabkin et al. JACC: CLINICAL ELECTROPHYSIOLOGY VOL. -, NO. -, 2017
New QTc formula - 2017:-–-

included in the QTcRTH conversion function training

F I G U R E 7 QTc for Each of the Different QTc Formulae in a Person Who Had 3 ECGs at
Different Heart Rates
set, whereas none of the cross-validated spline QTc
estimates were used to construct the conversion
function. By comparing the cross-validated spline
QTcs to the QTcs for competing formulae, we avoided
using the best fit conversion function for the whole
data as our reference. Doing so would result in overly
optimistic results in favor of the spline QTc. The cross-
validation method is useful for assessing the perfor-
mance of the spline approach. However, the entirety of
the NHANES data was used to construct the final spline
QTc conversion function; the spline conversion
formulae calculated in each fold of the cross-validation
were very similar (due to the large sample size) and
also very similar to the final spline-QTc.
The spline formula was not only good in samples of
the general populations but was also excellent when

The slope of the relationship between QTc and heart rate was calculated from linear
assessed under conditions in which many QTc
regression and was the smallest for QTc spline (0.015) and not significantly different formulae perform poorly. A sample of patient ECGs
(p ¼ 0.84) from zero. In contrast, there were positive slopes to the relationships that were was used to demonstrate the comparative value of the
significant for QTcBZT (slope ¼ 0.685; p < 0.001) and QTcDMT (slope ¼ 0.231; p ¼ 0.024) spline QTc. A large sample was not required, as the
but not for QTcHDG (slope ¼ 0.285, p ¼ 0.23). There were negative slopes to the QTc
objective was a comparative demonstration rather
heart rate relationship which were significant for QTcFRM (slope ¼ 0.454; p ¼ 0.019)
and QTcRTH (sex adjusted) (slope ¼ 0.085; p ¼ 0.03) but not for QTcFRD (slope ¼
than a methodologic evaluation or model develop-
0.139; p ¼ 0.0812). Abbreviations as in Figure 3. ment. The spline formulae provided a better heart rate-
corrected QTc in a sample of patients with heart rates
over 70 beats/min as shown by QTc values which were
not as prolonged as QTcBZT or as short as QTcFRD.
relationships remaining between heart rate and QTc
These data suggest that the widespread use of those 2
for the cross-validated spline QTc compared to most
formulae (QTcBZT and QTcFRD) should be reassessed.
other formulae.
The comparative data showed that the formula STUDY LIMITATIONS. A limitation of the study is the
which is most dependent on heart rate is that proposed use of interindividual data to define the QT-heart rate
by Bazett (4). This is apparent from the data consid- relationship rather than different heart rates in the
ering the QTc at the extremes of heart rate. The obvious same individual. Our data in a single individual,
differences between the QTcBZT and the relationship however, are illustrative of the stability of the spline
without a slope (the spline) show deviations of the formula across different heart rates, especially
overestimation of QTc at lower and higher QT in- compared to several other formulae. The power of the
tervals. This problem with QTcBZT has been noted cross-sectional data from NHANES is the large num-
previously (17). Many of the early formulae have been ber of individuals that were used to develop the
compared (21) but not the recent ones. QTcFRD also spline formula. It is challenging and likely not
failed at high heart rates. QTcFRM was reliable at possible to carry out a large study of the QT interval at
normal but failed at low and high heart rates. The for- different heart rates in the same individual at most
mula proposed by Hodges (13) was highly variable. institutions for several reasons. First, there are few
These deviations from a linear relationship with zero individuals with multiple ECGs in hospital, and those
slope for QTcBZT and QTcHDG are similar to the data with multiple ECGs are usually from patients in the
for the effect on QT interval with the increases in heart coronary care or intensive units whose serial ECGs
rate produced with exercise (20). show ST-segment and T-wave changes which would
The closest formula to the spline QTc was that pro- meet exclusion criteria for accurate QT measurement.
posed by Rautaharju et al. (9). We believe that this is Second, most other individuals with multiple ECGs
also an excellent formula. It used the entire NHANES often have the same heart rates in their different
database as well as several other data bases to obtain ECGs, thus precluding testing of the formulae across
parameters for their formulae. However, some of the meaningful different heart rates. Third, when an
QTcRTH performance may be slightly optimistic individual has a wide range of heart rates, in sinus
because it is likely that many of the individuals were rhythm, there are often only a few ECGs from which
JACC: CLINICAL ELECTROPHYSIOLOGY VOL. -, NO. -, 2017 Rabkin et al. 11
- 2017:-–- New QTc formula

to construct the QT-heart rate relationship. Recog- Furthermore, we used a rigorous statistical approach
nizing these challenges, further research with the to compare the spline formula to existing formulae
new spline formula should be conducted within the and found that the new formula was usually the best.
same individuals over a wide range of heart rates. We
note, however, that the QT correction formulae in REPRINT REQUESTS AND CORRESPONDENCE: Dr.
widespread use today, which were compared in our Simon W. Rabkin, Department of Medicine, Division
study, were constructed from patient samples or of Cardiology, University of British Columbia, Level 9,
populations and examined different heart rates in 2775 Laurel Street, Vancouver, British Columbia V5Z
different individual. In order to address this issue, 1M9, Canada. E-mail: rabkin@mail.ubc.ca.
our study developed the formula in one sample of the
population (cohort) and then tested it in different PERSPECTIVES
part of the cohort, the validation cohort. To our
knowledge, no other study deriving a widely used
COMPETENCY IN MEDICAL KNOWLEDGE: Current QT-heart
QTc formula has used this rigorous procedure.
rate correction (QTc) formulae do not satisfactorily accomplish
this objective especially at faster heart rates. A new QTc formula
CONCLUSIONS
based on functionally agnostic modeling of the NHANES popula-
tion ECG data using flexible regression splines overcomes this
A new approach, independent of the previous math-
problem. It further provides the data in percentiles of the popu-
ematical relationships, has been used to develop a QT
lation considering a person’s age and sex to permit improved
heart rate correction formula. The formula was
assessment of an individual’s probability of QT prolongation.
developed using one data set and tested on a separate
data set from the same large population. The formula
TRANSLATIONAL OUTLOOK: Research testing the formula in
is the closest theoretical formula to the goal of a for-
other populations and within individuals at different heart rates
mula with no relationship between QT and heart rate,
should be conducted. This QTc formula should be considered the
that is, no impact of heart rate on QTc. The new for-
new standard for adjustment of QT for heart rate.
mula is more “data agnostic” and more robust in its
construction than previous QT correction formulae.

REFERENCES

1. Roden DM. Drug-induced prolongation of the on rate correction formulas that minimize bias at the 17. Karjalainen J, Viitasalo M, Manttari M,
QT interval. N Engl J Med 2004;350:1013–22. upper normal limits. Int J Cardiol 2014;174:535–40. Manninen V. Relation between QT intervals and
heart rates from 40 to 120 beats/min in rest
2. Whitted A, Stanifer J, Dube P, et al. 10. Centers for Disease Control and Prevention.
electrocardiograms of men and a simple method
A dyshomeostasis of electrolytes and trace ele- National Health and Nutrition Examination Survey
to adjust QT interval values. J Am Coll Cardiol
ments in acute stressor states: impact on the II, electrocardiogram data. 2012. Available at: ftp://
1994;23:1547–53.
heart. Am J Med Sci 2010;340:48–53. ftp.cdc.gov/pub/Health_Statistics/NCHS/nhanes/
nhanes3/2A/nh2ecg.dat. Accessed January 2016. 18. Fox J. Nonparametric simple regression:
3. Bokil NJ, Baisden JM, Radford DJ, Summers KM.
smoothing scatterplots. In: Quantitative Applica-
Molecular genetics of long QT syndrome. Mol 11. Centers for Disease Control and Prevention.
tions in the Social Sciences. Thousand Oaks, Cali-
Genet Metab 2010;101:1–8. National Health and Nutrition Examination Survey
fornia: SAGE Publications; 2000:2.
III, electrocardiogram data. 2015. Available at:
4. Bazett H. An analysis of the time-relations of
ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/ 19. Funck-Brentano C, Jaillon P. Rate-corrected
electrocardiograms. Heart 1920;7:353–67.
nhanes/nhanes3/2A/NH3ECG-acc.pdf. Accessed QT interval: techniques and limitations. Am J
5. Fridericia LS. The duration of systole in an January 2016. Cardiol 1993;72:17B–22B.
electrocardiogram in normal humans and in pa- 12. Chang W, Cheng J, Allaire JJ, Xie Y, McPherson J. 20. Rabkin SW, Cheng X. Newer QT correction
tients with heart disease. 1920. Ann Noninvasive shiny: Web Application Framework for R. R package formulae to correct QT for heart rate changes
Electrocardiol 2003;8:343–51. version 0.13.2. Available at: https://CRAN.R-project. during exercise. Am J Med Sci 2016;351:133–9.
6. Rabkin SW, Cheng XB. Nomenclature, categori- org/package=shiny. Accessed May 20, 2016.
21. Puddu PE, Jouve R, Mariotti S, et al. Evaluation
zation and usage of formulae to adjust QT interval 13. Hodges M, Salerno D, Erlien D. Bazett’s QT of 10 QT prediction formulas in 881 middle-aged
for heart rate. World J Cardiol 2015;7:315–25. correction reviewed: Evidence that a linear QT men from the seven countries study: emphasis
7. Sagie A, Larson MG, Goldberg RJ, Bengtson JR, correction for heart rate is better. J Am Coll Car- on the cubic root Fridericia’s equation.
Levy D. An improved method for adjusting the QT diol 1983;1:1983. J Electrocardiol 1988;21:219–29.
interval for heart rate (the Framingham Heart 14. Lumley T. Analysis of complex survey samples.
Study). Am J Cardiol 1992;70:797–801. J Stat Softw 2014;9:1–19.
8. Dmitrienko AA, Sides GD, Winters KJ, et al. 15. Good P. Permutation, Parametric and Boot- KEY WORDS heart rate, QT interval,
Electrocardiogram reference ranges derived from strap Tests of Hypotheses. 3rd edition. New York QTc formulae
a standardized clinical trial population. Drug Inf J City: Springer; 2005.
2005;39:395–405.
16. Rabkin SW, Cheung B, Thompson DJS. Detailed
9. Rautaharju PM, Mason JW, Akiyama T. New age- analysis of the impact of age on the QT interval. A PP END IX For expanded equations, please
and sex-specific criteria for QT prolongation based J Geriatr Cardiol 2016;13:740–8. see the online version of this article.