Electrocardiographic Predictors of Coronary Heart Disease and Sudden Cardiac Deaths in Men and Women Free From Cardiovascular Disease in the Atherosclerosis Risk in Communities Study

Background We evaluated predictors of coronary heart disease (CHD) death and sudden cardiac death (SCD) in the Atherosclerosis Risk in Communities (ARIC) study. Methods and Results The study population included 13 621 men and women 45 to 65 years of age free from manifest cardiovascular disease at entry. Hazard ratios from Cox regression with 95% confidence intervals were computed for 18 dichotomized repolarization‐related ECG variables. The average follow‐up was 14 years. Independent predictors of CHD death in men were TaVR‐ and rate‐adjusted QTend (QTea), with a 2‐fold increased risk for both, and spatial angles between mean QRS and T vectors and between Tpeak (Tp) and normal R reference vectors [θ(Rm|Tm) and θ(Tp|Tref), respectively], with a >1.5‐fold increased risk for both. In women, independent predictors of the risk of CHD death were θ(Rm|Tm), with a 2‐fold increased risk for θ(Rm|Tm), and θ(Tp|Tref), with a 1.7‐fold increased risk. Independent predictors of SCD in men were θ(Tp|Tref) and QTea, with a 2‐fold increased risk, and θ(Tinit|Tterm), with a 1.6‐fold increased risk. In women, θ(Tinit|Tterm) was an independent predictor of SCD, with a >3‐fold increased risk, and θ(Rm|Tm) and TV1 were >2‐fold for both. Conclusions θ(Rm|Tm) and θ(Tp|Tref), reflecting different aspects of ventricular repolarization, were independent predictors of CHD death and SCD, and TaVR and TV1 were also independent predictors. The risk levels for independent predictors for both CHD death and SCD were stronger in women than in men, and QTea was a significant predictor in men but not in women.

E valuation of the risk of adverse cardiac effects for QT prolongation has been the focal point of numerous clinical and epidemiological studies. 1 However, QT is known to have notable limitations, [2][3][4] and professional organizations and governmental regulatory agencies have recognized the need for more sensitive predictors of adverse cardiac events than QT. [5][6][7] There is limited information about the utility of repolarization subintervals and associated repolarizationrelated ECG variables for prediction of adverse cardiac events. However, several investigations have found increased mortality risk in various clinical and general populations for the spatial angle between the mean QRS and T vectors [h(R m T m )]. [8][9][10][11][12] The objective of our study was to evaluate the risk of coronary heart disease (CHD) death and sudden cardiac death (SCD) in cardiovascular disease (CVD)-free men and women for a comprehensive set of repolarization-related ECG parameters.

Study Population
The Atherosclerosis Risk in Communities (ARIC) study was designed as a prospective investigation of the cause and natural history of atherosclerosis, its clinical manifestations, and the community burden of CHD. Risk factors were measured and outcomes evaluated in a population-based probability sample of adults 45 to 65 years of age at the 1987 to 1989 baseline examination; follow-up of the cohort is ongoing. Study population and definitions of prevalent diseases at baseline and outcomes have been described previously. [13][14][15] Deaths were classified into definite or possible CHD death, non-CHD death, and unclassified death.
SCD was defined as definite or possible CHD death that occurred within 1 hour after the onset of acute symptoms. CHD at baseline was classified by angina pectoris using the questionnaire of Rose et al 16 or myocardial infarct (MI), defined by a self-reported episode requiring hospitalization for >1 week, MI diagnosed by a physician, major Q waves at the baseline ECG (Minnesota Code 1.1), 17 or previous coronary artery bypass graft or coronary angioplasty. Prevalent (baseline) heart failure (HF) was determined on the basis of evidence of use of HF-related medications and classified according to Gothenburg criteria. 18 Baseline cerebrovascular disease was defined as self-reported stroke or transient ischemic attack that was verified by a study physician's review of the reported symptoms.
After exclusion of ECGs of participants with bundle branch blocks, artificial pacemakers, Wolf-Parkinson-White pattern, and technical errors in ECG recording detected in visual inspection of all the study ECGs using computer graphics terminals, source data for the present investigation were available from 15 005 ARIC participants. Participants with CVD at baseline (n=1384) were excluded from the present study (CHD, hospitalized HF, or cerebrovascular disease classified by criteria as noted above), leaving a CVD-free subgroup of 5937 men and 7684 women for the present study. The outcome data for CHD and SCD were available from a mean follow-up period of 14 years.

Electrocardiographic Procedures and Quality Control
Standardized procedures were used for recording the 12-lead ECGs with MAC Personal Computer (Marquette Electronics, Milwaukee, WI) in each clinical center. ECGs were processed in a central ECG laboratory initially using the Dalhousie Novacode ECG program. 19 All ECGs were later reprocessed with the GE Marquette 12-SL program (GE Marquette, Milwaukee, WI). The quasi-orthogonal XYZ leads were derived from the 8 linearly independent component-leads of the 12lead ECG signals using Kors' transformation, 20 and these leads were used as the source data to derive ECG parameters for the repolarization model. QTpeak (QT p ), QTend (QT e ), and QTonset (QT o ) intervals were first rate-adjusted as power functions of the RR interval derived in CVD-free men and women by regressing lnQT on lnRR. The exponent for RR was 0.33 for QT e for men and women and ranged from 0.36 to 0.40 for QT p and QT o . It was noticed that as long as proper functional form was used for rate adjustment, the exact value of the exponent for the RR interval had little influence on the R-square values. Additional analyses were performed to compare the above exponential rate adjustment formulas with simpler linear functions of the RR interval in the CVD-free groups of men and women ( Table 1). The results revealed that the differences in the accuracy of rate adjustment in terms of the R-square values were quite small. Recognizing that using different rate adjustment functions for QT e and QT p is an added complexity, a simpler QT p rate adjustment as the difference (QT ea -T p T e interval) was derived as shown in the middle section of Table 1. This became possible because the T p T e interval in the CVD-free groups was practically independent from heart rate (R-square, 0.120 for men and 0.045 for women).

Definitions of Repolarization Parameters
A set of 18 repolarization-related ECG variables was chosen for evaluation based on previous data of the variables' value as risk predictors or because of their functional role in the generation of normal and abnormal repolarization waveforms. QRS duration was included among these repolarizationrelated parameters because even moderate QRS prolongation has been shown to induce secondary repolarization abnormalities associated with adverse cardiac events. The conceptual model used to derive the repolarization parameters for the present study has been described in previous publications. 21,22 Temporal landmarks and measurement points for key intervals and amplitudes in the repolarization model are shown in the sketch of the ST-T vector magnitude curve in Figure 1 Table 3). T p -T xd , in turn, is the interval from T p to T xd , where T xd is the inflexion point (the minimum slope) at global T-wave downstroke. Left ventricular (LV) RT at point T xd (RT xd ) is obtained with an algorithm similar to that for RT p , whereby RT xd =QT pa +{[Cos h(T p |T ref )+1]9[T p T xd ]}/2. The key role of RT p and RT xd for deriving ECG estimates for RT epi and RT endo is considered in detail in the Discussion section in the subsection "Validity of the Repolarization Model." In addition to h(T p |T ref ), a number of other spatial angles reflecting deviations of the direction of repolarization from the reference normal direction during various repolarization subintervals and other repolarization-related interval and amplitude variables were used in various phases of the study. Their definitions are listed in the footnotes of the corresponding tables.

Statistical Methods
One baseline ECG per participant was used for all analyses. Mean values and standard deviations were determined for continuous variables and frequencies and percents for categorical variables. Cox proportional hazards regression was used to assess associations of ECG variables with the risk of CHD death and SCD using both univariable and multivariable risk models. Predictor ECG parameters were first evaluated as continuous variables and then dichotomized using quintiles to define test and reference groups. The thresholds for test groups are listed in Table 3. Hazard ratios (HRs) were evaluated for increased values of the ECG parameters (quintile 5) as the test group, with quintiles 1 to 4 as the reference group. However, quintile 1 corresponding to decreased values was used as the test group for TaVL and T p V, with the remaining 4 quintiles as the reference group. ECG predictors were first evaluated as unadjusted single variables and subsequently in multivariable-adjusted models with adjustment for age, sex, center, race, education, smoking status, diabetes, hypertension, family history of CHD/stroke, BMI, SBP, ratio of total cholesterol/highdensity lipoprotein, glucose, creatinine, and uric acid. An association was considered significant when P<0.05 and no adjustment for multiplicity of comparisons had been considered. Finally, to identify independent ECG risk predictors, those ECG variables that were significant predictors in single variable models were entered simultaneously into the Cox regression model, and each was adjusted to other ECG variables.
Participants who had no events during the follow-up period were censored in the analysis at their date of last contact. A participant who died from CHD with the death not SCD was censored at the date of the CHD non-SCD death. No attempt was made to evaluate possible competing risks. All analyses were performed with SAS version 9.1.3 (SAS Institute, Cary, NC).

Study Group Characteristics
Demographic and clinical characteristics of the study population listed in Table 2 have been described in more detail in previous publications. 21, 23 The age range of the study population was 45 to 65 years, with mean age of 54 years (SD, 5.7 years). The study population was predominantly white (73%). The prevalence of hypertension was %30% in men and women.
ECG variables including all repolarization-related parameters evaluated are listed in Table 3. Most of the differences between men and women were statistically significant. QT pa , the rate-adjusted QTpeak, was 18 ms shorter and QT oa , the rate-adjusted QTonset, 21 ms shorter in men than in women. The sex difference in global rate-adjusted QT (QT ea ) was smaller, 10 ms. h(R m |T m ), the spatial angle between the mean QRS and T vectors, was 12°wider in men than in women. Among other notable differences in ECG parameters, T o V, Tonset vector magnitude, was %30% lower in women than in men, and a similar difference was observed in T p V, the spatial magnitude of the Tpeak vector.

CHD Death Predictors
Summary results are presented Table 4 for ECG variables evaluated in Cox regression as unadjusted single predictors and as multivariable adjusted single ECG predictors. From the set of 18 ECG variables, 12 in men and 13 in women were significant predictors of CHD death in unadjusted single ECG variable models. The set of significant unadjusted single predictors was the same in men and in women except that QRS duration was a significant predictor in women but not in men. Angular variables and T-wave amplitudes were the strongest single predictors of CHD mortality risk. Seven of the predictors in men and 5 in women remained significant after adjustment for demographic and clinical risk factors.

Sudden Cardiac Death Predictors
As for CHD death, many repolarization-related variables (11 in men and 11 in women) were significant predictors of SCD when evaluated as unadjusted single ECG-variables (Table 5). Significant predictors were in general the same parameters in both sexes except that QT ea and the rate-adjusted T p T e interval ([T p T e ] a ) were significant predictors in men but not in women. HRs were particularly high for some of the ECG predictors in women: in 9 of them, HR was >2-fold and in 7 of them >3-fold. Five predictors in both men and women remained significant in the fully adjusted multivariable model.

Independent Predictors of CHD Death and SCD
Angular variables were commonly chosen as independent predictors of CHD death and SCD (Table 6). In addition, QT ea was a significant independent predictor of CHD death and SCD in men, TV1 an independent predictor of CHD death and SCD in women, and TaVR an independent predictor of CHD death in men. Risk levels in women for both CHD death and SCD were stronger than in men. In terms of the magnitude of increased risk of CHD death for these ECG predictors, in men TaVR and QT ea were the strongest predictors, with a 2-fold increased risk for both variables. Also significant independent predictors were h(R m |T m ) and h(T init |T term ) with an >1.5-fold increased risk (although the P values were marginally significant). In women, the risk of CHD death was increased 2-fold for h(R m |T m ) and increasd 1.7-fold for h(T p |T ref ).
Independent predictors of SCD in men were h(T p |T ref ) and QT ea , with a 2-fold increased risk, and h(T init |T term ), with a 1.6-fold increased risk. In women the strongest independent predictor of SCD was h(T p |T ref ) (HR, 3.55; CI, 1.85 to 6.81; P<0.001). In addition, the risk of SCD for h(R m |T m ) and TV1 was increased >2-fold for both.
In terms of sex differences, the risk levels in women for both CHD death and SCD were stronger than in men. Another notable sex difference was that QT ea was a significant independent predictor of CHD death and SCD in men but not in women.

Discussion
A majority of the 18 ECG parameters were significant CHD death and SCD predictors when evaluated as unadjusted single ECG variables, and many remained significant in multivariableadjusted models. Notable among these predictors were h(R m | T m ), the spatial angle between the mean QRS and T vectors, and h(T p |T ref ), the spatial angle between Tpeak and the normal T reference vector. h(R m |T m ) is a measure of the overall deviation angle between depolarization and repolarization sequences, and h(T p |T ref ) is a measure of deviation of the direction of the repolarization sequence from the normal reference direction during regional cross-mural repolarization of the left ventricular lateral wall. h(R m |T m ) was also an independent predictor for CHD death in men, with a 62% increased risk, and in women, with a 2-fold increased risk, and h(T p |T ref ) was a strong  Journal of the American Heart Association independent predictor for SCD, with a nearly 2-fold increased risk in men and a >3-fold increased risk in women. It is worth noting that increased CHD death and SCD risk were observed in 20% of men and women (the upper quintile), with a relatively moderate widening (23°to 30°) of h(T p |T ref ) and h(T init |T term ). QT prolongation was an independent predictor for CHD death and SCD in men but not in women.

Validity of the Repolarization Model
As noted in the Methods section, RT p is computed as a function of the QT pa , which is the key parameter in the algorithms for RT p and RT xd . RT p is conceived by the repolarization model to represent the RT of LV myocytes, which repolarize at the time of T p during the initial fast phase of LV lateral wall repolarization. It is noted from the algorithms for deriving RT p as a function of QT pa that the RT p is modified by the cosine of h(T p |T ref . These functional relationships in the repolarization model are based on consideration of potential theory as applied to the generation of T wave, and they differ from the notions from some electrophysiological reports on animal models using wedge preparations that T p timing always coincides with QT epi . [23][24][25][26] Potential theory supports the assertion that at the time of RT p the majority of LV lateral wall myocytes are in phase 3 of their action potential and that RT p timing coincides with the timing of the global T p. It thus seems a rational proposition to maintain the labels RT epi and RT xd as derived by modifying RT p by h(T p |T ref ). In a strict sense, the label RT epi refers to the RT of subepicardial myocyte layers and should be considered as a representative value of LV epicardial RT and not RT at any specific epicardial location. Potential theory is also compatible with the notion that at the time of T p and RT p , LV lateral wall subepicardial myocytes in the region where repolarization starts earliest have already reached phase 4 of their action potential, no longer contribute to the generation of the T wave, and T amplitude starts to decline. This occurrence is the likely explanation for electrophysiological data relating the timing of QT p with QT epi in normal repolarization. [23][24][25][26] Parameter T xd in the repolarization model is the inflexion point (the minimum slope) at the global T-wave downstroke, considered by the repolarization model to occur when the largest number of LV myocytes leaves phase 3 of their action potential within the same increment of RT. With the normal direction of the RT sequence, this conceivably occurs when the majority of subendocardial myocytes reach their resting potential. Spatial direction of repolarization during the T p T xd interval is diametrically opposite to the direction of the T p vector, and T p T xd is the magnitude of the temporal RT gradient vector representing RT dispersion during the T p T xd interval dominated by the LV lateral wall repolarization. Contrary to the notions from electrophysiological studies suggesting that LV lateral wall repolarization is perpendicular to the epicardial surface, [23][24][25] there is consistent evidence that the spatial LV repolarization sequence remains throughout repolarization closely in the direction from inferior-left-anterior to superior-right-posterior, approximately in the direction of the lead vector of aVR but with a posterior component. 21,22 This implies that LV lateral wall repolarization is cross-mural, oblique rather than perpendicular to the epicardial surface. 10,21,22,27

Mechanisms of Generation of Repolarization Abnormalities as Independent Predictors of CHD Death and SCD
Anterior-right rotation of the T p vector is a predominant determinant of widened h(R m |T m ) and h(T p |T ref ) angles. 10,21 T p vector rotation closer to the aVR lead axis results in decreased (less negative) TaVR amplitude that ultimately becomes positive with a more pronounced widening of h(T p | T ref ). Altered direction of the repolarization sequence may reflect subepicardial action potential duration shortening such as takes place in anterior subepicardial myocardial ischemia. 21 Thus, the increased h(T p |T ref ) observed in the present study as a common predictor for CHD death and SCD may possibly be an early marker of evolving subclinical CHD in men and women free from manifest CVD. h(T p |T ref ) widening also reflects a gradual change from the normal predominantly reverse sequence of the cross-mural left ventricular wall repolarization to a concordant repolarization with respect to depolarization and increasing dyssynchrony of repolarization 10 that in turn has been postulated to be associated with increased dyssynchrony of ventricular repolarization as another possible risk mechanism. 26 QT ea was not an independent predictor for either end point in women, but in men it was a significant independent predictor, with a 48% increased risk for CHD death and a 98% increased risk for SCD. Sex difference in QT is actually not a result of QT prolongation in women, as commonly claimed, but arises from pronounced QT shortening in adolescent boys. 28 QT gradually prolongs with age in adult men, and the sex difference becomes small or vanishes after middle age.  from cardiovascular disease highlighting sex differences in rateadjusted QTpeak (QT pa ), QTend (QT ea ), and QTonset (QT oa ) intervals and Tpeak (T p ) amplitudes. The ST-T curve was generated by sampling the ST-T vector magnitude function between the end of QRS and the end of the T wave (T e ) at 60 equally spaced sample points, subsequently aligning the Tpeak (T p ) time with the mean QT pa and rescaling the temporal RT axis back to the original to match T p T e and T o T p intervals with the mean QT pa and T oa -T pa intervals of men and women. Compared with that in women, repolarization in men starts earlier, the ST segment has a steeper upslope, and T p is shifted to the left, corresponding to an earlier end of epicardial repolarization as reflected by an 18-ms shorter QT pa in men. CVD indicates cardiovascular disease.
Although QT ea is a measure of the global RT, QT pa and RT epi are measures of regional RT. The present investigation revealed an 18-ms sex difference in QT pa (Figure 2), indicating that the sex difference in RT p remains more pronounced in middle-aged men and women than the 10-ms sex difference in QT ea as listed in Table 3. It is not known whether prolonged regional repolarization time (RT p ) might play some role in explaining the higher vulnerability of women than men to the proarrhythmic effects of cardioactive drugs. 29

Comparison With Previous Studies
Two recent publications in general population samples of men and women 30,31 and 1 in men 32 found QRS duration to be predictive of SCD. In our study population of CVD-free men and women, QRS duration was a significant predictor only in the unadjusted single ECG variable risk model in women for CHD death. Several publications have documented an increased mortality risk for a wide mean QRS|T angle [8][9][10][11][12] and abnormal T-wave axis. 33,34 Various angular measures of altered repolarization sequence were the most common predictors for CHD death and SCD in our study. Increased T-wave amplitude in aVR was reported to be a significant predictor of cardiac mortality in the general population of men and women 35 and in a large clinical male population. 36 The previous investigations cited above have evaluated ECG predictors of CHD death and SCD as single variables or as a limited group of variables. The present investigation is the first large-scale population study with simultaneous evaluation of a comprehensive set of repolarization-related parameters.

Limitations of This Investigation
Although the multivariable models employed were adjusted for a variety of demographic and clinical factors, competing risk analysis was not performed. The primary objective of the study was to identify in CVD-free men and women a subset of ECG parameters for future risk evaluation studies as potentially more sensitive predictors of CHD death and SCD than the QT interval.
Clinical Significance and Avenues for Future Research h(R m |T m ) and h(T p |T ref ), reflecting different aspects of ventricular repolarization, were found to be independent predictors of CHD death and SCD, and TaVR and TV1, readily available in standard ECG reports, were also independent predictors. Among notable sex differences, the risk levels for independent predictors for both CHD death and SCD were stronger in women than in men, and QT ea was a significant predictor in men but not in women. These ECG variables identified as independent predictors of CHD death and SCD are the primary candidates that warrant consideration in risk evaluation studies. However, all the repolarization-related parameters that were significant when evaluated as single variables need attention in the evaluation of possible markers of toxic drug effects using wellvalidated annotated data files from drug trials.