Академический Документы
Профессиональный Документы
Культура Документы
P
rolongation of the QT interval is associated with
Several form&s have been proposed to adjust increased risk for malignant ventricular arrhyth-
the QT Interval for heart rate, the most commonly mias and sudden cardiac death in congenital long
used bslng the QT correctlon formula (QT, = QT syndromes, electrolyte disturbances, after use of
QT/m) propwed In 1920 by Raxett. The QT, several antiarrhythmic drugs and after myocardial in-
formula was derived from observations In only 39 farction.-* The duration of the QT interval has been
young subjects. Recently, ths adequacy of Ba- shown to depend on the length of the precedmg cardiac
&ts formula has been questtoned. To evaluate cycle. It has been over 70 years since a formula for cor-
ths heart rate QT assocMon, the QT Interval was recting QT for cycle length (QTJ was introduced by
measured on the hdtlal baseline electrocardiogram Bazett.13 However, the adequacy of thii approach (di-
of S,OlS subjects (2,229 men and 2,779 women) viding the measured QT by the square root of the mea-
from the Framlngham Heart Study with a mean sured RR) to adjust correctly for differences in heart
age of 44 years (range 26 to 62). Persons wlth rate has been questioned. i4-*0 Although many formulas
coronary artery disease were excluded. A linear have been proposed to adjust the QT interval for heart
regmdon modei was developed for correcting QT rate, only a few have been derived or validated in a
according to RR cycle length. The large sample al- large population-based cohort.14-22 In the present study
lowed for subdlvlslon of the populatlon into sex- we examined the initial baseline electrocardiogram in
speclflc deciles of RR Intervals and for comparison over 5,000 persons originally enrolled in the Framing-
of QT, Raxetts QT, and linear corrected QT ham Heart Study. This report examines the relation of
(QT&. The mean RR interval was 0.81 second QT (and QT,) to heart rate and presents results of a
(range 0.5 to 1.47) heart rate 74 bsats/mln (range linear regression equation (QTLc) that better corrects
41 to 120), and mean QT was 0.35 second (range QT for heart rate.
0.24 to 0.49) In men and 0.36 second (range 0.26
to 0.46) In women. The linear regression model METHODS
yleldedacorrectlonformula(fora&erenceRR The original cohort of the Framingham Heart Study
Interval of 1 second): QTu: = QT + 0.154 (1 - RR) consisted of 5,209 subjects who were between the ages
that applies for men and women. This equatlon of 28 and 62 years on entering the study >40 years ago.
corrects QT more reliably than the Razetts formu- The initial electrocardiograms of these subjects were
la, which overcorrects the QT Interval at fast heart used to measure the duration of the QT interval. From
rates and undercorrects lt at low heart rates. Low- 1985 to 1986 all initial electrocardiograms were mea-
er and upper limits of normal QT values In relation sured manually by 2 trained readers with the aid of cali-
to RR were generated. A simple linear equation pers and magnifying lens. Measurements were obtained
was developed that Is mere accurate than Baxetts from a 1Zlead resting electrocardiogram, with a paper
correction at dlfferent cycle lengths and more con- speed of 25 mm/s. The average of 2 to 3 QT intervals
venient for cllnkal practke. This formula allevl- from the electrocardiographic lead with the longest QT
ates the need to apply sscondary corrections to interval was analyzed. The QT interval was measured
9axetts fomuda. Addltlonal Investlgatlon Is war- from the beginning of the QRS complex to the end of
ranted to determined whether QTLC Improves the the T wave where its terminal limb joined the baseline.
ldentlftcation of subjects at high risk for mallg- Each week a blinded duplicate review was performed on
nant arrhythmias or sudden death. a random sample of all electrocardiograms read during
(Am J Cardlol1992;70:797-601) that week. There was a high degree of inter- and in-
traobserver agreement. Any disagreements found were
From the Framingham Heart Study, Framingham, the Divisions of subsequently adjudicated and corrected. Additionally, a
Epidemiology and Preventive Medicine of Boston University School of sample of 480 baseline electrocardiograms were blindly
Medicine, Boston, the Department of Medicine, University of Massa-
chusetts Medical School, Worcester, and the Divisions of Cardiology and independently reread 5 times. A high degree of cor-
and Clinical Epidemiology, Beth Israel Hospital, Boston, Massachu- relation was obtained between the multiply read sam-
setts; and the National Heart, Lung, and Blood Institute, Bethesda, ples and the singly read baseline measurements with re-
Maryland. Dr. Sagie was a visiting Research Fellow from the Beiin gard to the mean RR, and average QT interval (p
Medical Center, Tel Aviv, Israel. Manuscript received May 1, 1992;
revised manuscript received and accepted May 21,1992. <O.OOl). Subjects with coronary heart disease (n = 82)
Address for reprints: Daniel Levy, MD, Framingham Heart Study, at the initial examination were excluded from consider-
5 Thurber Street, Framingham, Massachusetts 01701. ation because coronary disease may have affected the
0.42 T
0.40
0
v Men: actual
- -Men: Linear
0 Women: actual
-Women: Linear
0.32
RR, RMSE = 0.0224 and R squared = 0.445 for wom- The basic linearity and goodness-of-fit for the QT-
en). Because of these similarities men and women were RR relation is demonstrated in Figure 1, where each
pooled for further analyses. data point represents mean QT versus RR decks. This
Stepwise regression was used to investigate the si- plot establishes that the fitted lines adhere closely to the
multaneous effects on QT from RR, sex and age and group means with no evidence of gross departure from
their higher-order interactions, as well as RR2 to allow linearity at any RR value either in men or women.
for curvature. The single most important variable was QT eorre&ionr hear versus Bazettr The linearly
the duration of the RR cycle, which explained 42.2% of corrected QT (QT&, has the same expression for men
the QT variance; this was followed by sex which ex- and women, namely: QTLC = QT + 0.154 . (1.00 -
plained an additional 3.6% of the variance. Although RR) where QT and RR are measured values and 0.154
age, RR2, and the interaction term RR . age also were is the estimated regression slope. A subject with a heart
statistically significant (each at p <O.OOS), their clinical rate >60 beats/mm (RR <l) has a QTLC >QT, where-
and biologic importance were minimal, because none of as one with a heart rate <60 beats/min (RR >l) has a
these individually accounted for even 1% of the QT QTLC <QT (i.e., the linear correction is upward when
variance; all together they accounted for only 1.4% after the RR interval is shorter than 1 second and downward
adjustment for RR and sex. Therefore, we selected the when it is longer than 1 second).
2-variable linear model: Fitted QT = 0.234 + 0.154 . Mean QT, and QTLC were computed for each RR
RR - 0.012 - male. This model yielded an RMSE of decile (Figure 2). The mean QTLC values are virtually
0.0224 and R2 of 0.458, and was the simplest model constant across the range of RR values. In contrast the
that was adequate to predict QT in our combined sam- QTC values decline with increasing RR decks. Thus,
ple of men and women. the linear correction for QT yields a valid correction;
0.42
\ \
\ , , Women QT c
MenQT;L-,
-. 1---
--. \ \
\
\ --- \
I I I I I
RR Interval (set)
the group mean QTLC values were between 0.313 and tion of this interval is associated with increased inci-
0.379 for men, and between 0.384 and 0.391 for wom- dence of malignant ventricular arrhythmias and sudden
en. Small sex differences in QTLC were observed with death.-I2 To use heart rate-corrected QT to predict
mean values of 0.376 in men and 0.388 in women. Be- life-threatening arrhythmias and sudden death, it would
cause of the large sample size, limits for the central 95% be appropriate to use a QT correction formula that is
of individual QTLC values are approximated by the ex- developed and validated in a large population-based co-
pression QTLC (mean) f 1.96 . RMSE. These 95% lim- hort. Using Framingham Heart Study subjects, we de-
its for QTLC are 0.332 to 0.420 for a man and 0.344 to veloped a linear regression equation (QTLc) that more
0.432 for a woman. A subject whose QTLC is outside accurately corrects QT for heart rate than the tradition-
these sex-specific limits is considered to have an unusu- ally used Bazetts (QT,) formula. We found that QTc
ally short (or long) QT interval. Lower and upper limits undercorrects QT at slow heart rates and overcorrects it
of the QTC distribution are not constant over the ob- at fast heart rates, whereas QTLC reliably corrects QT
served range of RR, but must be determined anew at across a wide range of RR cycle lengths. The large
each RR value. The impact of different decks of RR number of subjects studied also enabled us to subdivide
intervals on QT, QTC and QTLC in men and women is the sample into deciles of RR intervals and to examine
presented in Tables II and III. the impact of different cycle lengths on QT, QTC and
Bazetts formula, QTC = QT/fl, also adjusts up QTLc.
ward when RR is <l, and downward when RR is > 1 Several formulas have been proposed to adjust the
second. But, it yields substantial overadjustment at QT interval for heart rate.13-22 The most frequently
short RR intervals and underadjustment at long RR used is Bazetts square root formula which was intro-
intervals (Figure 2). Bazetts correction is flawed, duced more than 70 years ago.13 However, the adequa-
whereas the linear adjustment QTlc produces a common cy of this nonlinear formula, obtained from data on 39
reference value for all members of the same sex. young men, has been questioned because the QTC over-
Another way to use the linear model for cliical pur- corrects the measured QT interval at fast heart rate and
poses is to construct 95% confidence limits for an indi- undercorrects it at a low heart rate.16-22 Our data con-
vidual QT observation at different RR cycle lengths. firm this deficiency of Bazetts correction. Puddu et all7
These data are listed in Table IV. A subject whose mea- examined 10 mathematically different QT prediction
sured QT falls outside the appropriate knits has an un- formulas in 881 middle-aged men and provided similar
usually short (or long) duration QT interval. This table evidence that the QT is overcorrected by Bazetts for-
alleviates the need to compute QTLc, although it may mula at high heart rates. They observed that the cubed
require linear interpolation when a measured RR lies root Fridericias formula20 fitted the data better than
between tabulated values. Bazetts or other formulas. In the present study we test-
ed the regression models used by Puddu and found that
DISCUSSION most of them (except Bazetts formula) adequately and
Assessment of a heart rate-adjusted QT interval is equally fitted the QT-RR data. Our results generally
considered to be clinically important, since prolonga- agree with those of Puddu,l but we included women as