The heart rate-corrected QT (QTc) interval in the resting 12-lead electrocardiogram (ECG) provides important prognostic information in clinical practice. In post-myocardial infarction patients (1), subjects referred for Holter monitoring (2), and subjects with diabetes (3) QTc prolongation is associated with mortality risk. In inheritable syndromes, which are characterized by QTc prolongation, affected patients have high risk of sudden death (4). Also, in several population studies (5- 10), increased risk of cardiovascular disease (CVD) mortality was observed in subjects with QTc prolongation. Less is known about the association with incident coronary heart disease (CHD). Furthermore, though racial differences in QT interval have been reported (11), there is no information about the association of QTc with CVD mortality or incident CHD in the black population. Therefore, we studied QT in relation to risk of CHD incidence and CVD mortality in a large biracial U.S. population study, the Atherosclerosis Risk In Communities (ARIC) study.

Because measurement of QT is hampered by systematic differences between manual coders and between computerized QT measurements of different ECG systems (12), manually determined QT intervals were compared with machine-derived QT intervals. In addition, to compare QT intervals between subjects, adjustment for differences in heart rate is essential. The widely used Bazett formula is known to overadjust at very high and low heart rates, and several alternative methods have been proposed (13- 14). We compared three commonly used formulae for the correction of heart rate.

The ARIC study is following a cohort of 15,792 middle-aged men and women (15). During 1987 to 1989, population samples were drawn from the 45- to 64-year-old inhabitants of Forsyth County, North Carolina; Jackson, Mississippi; the northwest suburbs of Minneapolis, Minnesota; and Washington County, Maryland; they were invited for a home interview and a clinical examination. In Jackson, 46% of eligible subjects participated; approximately 65% participated in the other three centers. For the present study, subjects with a self-reported history of physician-diagnosed heart attack, prior cardiovascular surgery or coronary angioplasty, or with 12-lead Minnesota Code evidence of prior myocardial infarction were excluded, leaving 14,672 subjects. After exclusion of subjects with missing data on QT, the study population consists of 14,548 men and women.

In the home interview, questions were asked about health behaviors, sociodemographics, and disease history. The clinical examination included electrocardiography. After resting 5 to 10 min while electrodes were placed, a standard 12-lead ECG and a 2-min paper recording of a three-lead (leads V₁, II, and V₅) rhythm strip were made in the supine position. The QT interval from the digital 12-lead ECG was determined by the NOVACODE program (16). The NOVACODE program generates an average waveform derived from all 12 simultaneously measured leads. This allows the system to determine the QT from the earliest QRS onset to the latest offset of the T-wave. All NOVACODE measurements were visually verified. When measurements between computer and operator differed by >5 ms, the operator's reference points were used. Plasma total cholesterol and triglycerides were measured by enzymatic methods (17- 18), and low-density lipoprotein cholesterol was calculated using the Friedewald formula. High-density lipoprotein cholesterol was measured after dextran-magnesium precipitation of non–high-density lipoproteins (19). Serum insulin was measured using a radioimmunoassay (¹²⁵I Insulin Kit, Cambridge Medical Diagnostics, Billerica, Massachusetts). Serum glucose was assessed by the hexokinase method. Prevalent diabetes mellitus was defined as a fasting glucose level of 126 mg/dl (7.0 mmol/l) or more, nonfasting glucose level of 200 mg/dl (11.1 mmol/l) or more, and/or a history of, or treatment for, diabetes. Three seated blood pressure measurements were taken on the right arm, using a random-zero sphygmomanometer. The mean of the last two measurements was used. Hypertension was defined as systolic blood pressure of 140 mm Hg or more, diastolic pressure of 90 mm Hg or more, or use of antihypertensive agents. The ratio of waist (umbilical level) and hip (maximum buttocks) circumferences was calculated as a measure of fat distribution. Average carotid intima-media thickness was assessed using a standardized B-mode ultrasonic technique (20). Use of cardiac medication was defined as self-reported use of medication for angina pectoris, heart failure, or rhythm disturbances during the last two weeks.

The method of follow-up has been described previously (21). Briefly, interviewers contacted participants annually by telephone to identify hospitalizations and deaths. Death certificates and discharge lists from local hospitals were surveyed by ARIC staff to detect additional deaths and cardiovascular events. For hospitalized patients with potential acute CHD events, trained abstractors recorded the presenting signs and symptoms, including chest pain, cardiac enzymes, and related clinical information. Up to three 12-lead ECGs were visually coded using the Minnesota Code, and wave-form evolution was evaluated using side-by-side comparisons. Out-of-hospital deaths were investigated by means of the death certificate and, in most cases, an interview with next of kin and questionnaires completed by the patients' physicians. Coroner reports and autopsy reports, when available, were used in validation. All potential in- and out-of-hospital clinical CHD events were reviewed and adjudicated by the ARIC Morbidity and Mortality Classification Committee using published criteria (21).

In the present study, follow-up data through 1993 were used. Follow-up information on incident CHD and mortality was complete for 95%. The CHD incidence was defined as definite or probable myocardial infarction, cardiac revascularization procedures (excluding thrombolytic therapy), or definite CHD death. Cardiovascular disease mortality was based on the underlying cause from the death certificates: ICD-9 codes 390 to 459.

QT was manually determined in a case-cohort sample from the ARIC study (21), consisting of a random sample of 900 subjects, all incident CHD cases, and all deaths. Measurements were performed at the University of Minnesota ECG Coding Center, using a digitizing tablet (Calcomp, Columbia, Maryland) and a personal computer. Coders mounted the 3-lead two-min rhythm strip on the tablet, and used the mouse to mark QT onset and T-wave offset and preceding RR intervals in three consecutive normal beats in all three leads. T-wave offset was defined as the point of maximum change of slope as the T-wave merges with the baseline (6). This accords with the method applied in the NOVACODE program (16). The inter- and intracoder variability was evaluated by means of analysis of variance in a set of 10 ECGs, which were measured three times by each coder. The QT interval was used as the dependent variable. Between-subject variance was 356 ms², the between-coder variance 79 ms², and the intercoder variance 15 ms².

The QT was corrected for heart rate using three different methods: Bazett (22) , the linear method proposed by Hodges et al. (14) (QThodges=QT+1.75[heart rate−60] [QT in ms]) , and by the method proposed by Rautaharju (13) (QTindex=QT · [heart rate+100]/656 [QT in ms]). Age-, gender-, and race-adjusted baseline characteristics were compared over categories of gender-specific quintiles of QTc, QThodges, and QTindex. Age-, race-, and gender-adjusted baseline characteristics were derived from regression analyses, using the specific characteristic as the dependent, and age, gender, and QT categories as explanatory variables. Relative risk of cardiovascular disease mortality, CHD incidence, and total mortality in gender-specific quintiles and in the upper 10% of QTc, of QThodges, and of QTindex were estimated using Cox proportional hazards analysis in the complete cohort. In the multivariable models, we forced in risk factors, which were previously suggested, as possibly confounding or mediating in the literature. For analyses of the manually determined QT, Poisson regression for the case-cohort design was used (23).

The correlation between the three QTc values was high: 0.81 between QTc and QThodges, 0.96 between QTc and QTindex, and 0.96 between QThodges and QTindex. Mean QTc was 436 ± 25 ms in black women, 435 ± 23 ms in white women, 421 ± 27 ms in black men, and 422 ± 22 ms in white men. In (Table le1), age-, gender-, and race-adjusted baseline characteristics are shown across gender-specific quintiles of QTc. In all groups, men and women, and in blacks and whites, a more adverse cardiovascular risk profile with increasing QTc was observed. Intima-media thickness and the prevalence of hypertension, diabetes, and electrocardiographic abnormalities were greater with increasing QTc. Longer QTc was also associated with lower serum potassium and calcium concentrations. Similar results were observed in categories of QThodges and QTindex (data not shown).

In (Table le2), relative rates are shown in quintiles of QTc. Risk of CVD mortality, incident CHD, and total mortality was greater with longer QTc in black men and women and in white women. In white men, a U-shaped association was observed for CVD mortality. In the total cohort, the age-, gender-, and race-adjusted hazard ratios of CVD mortality, but not of incident CHD, in the lowest quintile were higher than in the second quintile. From the fourth quintile on, relative risk for all end points increased. When adjusting for other cardiovascular risk factors one by one, none of these explained a large part of the increased risk. After adjustment for all other risk factors combined, only risk in the highest quintile was still significantly elevated. The large increase in risk in the highest quintile was mostly explained by the 10% with the longest QT interval. The age-, gender-, and race-adjusted hazard ratios for CVD mortality and CHD incidence were about 4 and 2, decreasing to about 2 and 1.5 after adjustment for all other risk factors (Table le3). Results were similar for QTc, QThodges, and QTindex. Stratification for race showed higher relative risks among the black than among the white population (Table le3). A test of statistical significance showed significant interaction between QT prolongation and race. Using race- and gender-specific cutoff points did not change the observed difference (data not shown). The reduction of the prognostic value of QTc prolongation in the multivariable model in the white population was observed in the women only. This may be due to low event rates because none of the individual variables explained this. The effect was observed only when all variables were included, and it disappeared when a lower cutoff point was used.

The results did not differ over strata of heart rate, diabetes, hypertension, or presence of ECG abnormalities. When stratifying for the presence of diabetes, hypertension, use of cardiac medication, or ECG abnormalities combined, the predictive value (after adjustment for other risk factors) of QTc and QTindex was somewhat higher in the healthy subpopulation (Table le4).

The average machine-determined QT interval was 10 ms longer than the manually determined QT, and the correlation was 0.81. In the age- gender-, and coder-adjusted model, the predictive value of the manual QTc was somewhat lower than that of the machine QT. However, this was due to the higher correlation of the machine QT with heart rate; the Pearson correlations of manual- and machine-coded QTc with heart rate were 0.33 and 0.48, respectively. After adjustment for heart rate, they had similar prognostic value (Table le5).

In this population-based cohort study of middle-aged men and women, a longer QTc was associated with an adverse cardiovascular risk profile. Prolonged QTc was a strong predictor of incident CHD and especially of CVD mortality, independent of the other risk factors. The association was stronger in black than in white subjects.

QT was determined in the digital ECG, and manually in the three-lead rhythm strip, which were recorded at the same visit. We observed a high agreement between machine- and manual-determined QT intervals, and both had similar prognostic value. Therefore, it can be concluded that NOVACODE-determined QT intervals can safely be used in population studies. This may not be true for patient populations, because the normal pattern is easier for automatic measurements to detect (14).

The QT interval has to be corrected for heart rate to compare individuals with different heart rates. There are several methods, Bazett's being the most commonly used. As illustrated by our data, most methods will be highly correlated in a normal population. The Bazett correction has been criticized for the residual correlation with heart rate. However, this does not explain the association with CVD risk. In the present study adding heart rate to the age- and gender-adjusted model, the relative rates of CVD mortality for QTc, QTindex, and QThodges were 4.37, 3.29, and 3.73, respectively. The prognostic value of QTc was slightly better than QTindex or QThodges.

Certain drugs are known to lead to QTc prolongation, and may introduce bias. However, exclusion of subjects using antihypertensive or cardioactive drugs did not change the observed associations in our study, and the associations were also present in subjects without signs of heart disease or chronic diseases.

This study is in line with previous reports on the association of prolonged QTc with mortality risk (5- 10) and of CHD incidence (6,8). The finding of a gradually increasing risk of CVD mortality with increasing QTc, and a nonlinear increase in the upper 10% of the population distribution was also observed in the Strong Heart study of American Indians (10). We showed the QT-associated risk may be even higher in the black population. This may be a real phenomenon because there are known racial differences in the 12-lead ECG, and race-specific criteria for left ventricular hypertrophy have been proposed (11,24). In healthy subjects without hypertension, diabetes, or history of cardiovascular disease in the ARIC study, black subjects had higher QRS voltage than white subjects, and QTc was shorter, especially after adjusting for other cardiovascular risk factors (11). When we increased the cut point for prolongation for the white subjects in the present study, the relative risks in white subjects increased somewhat, but they were still considerably lower than in black subjects. Furthermore, racial differences have also been reported in the mechanism of CHD leading to mortality (25). The observed higher risk may contribute to the higher case fatality in the black population (26). The interaction with race may also be a chance finding because previous studies reported higher relative risks in white subjects with prolonged QTc than observed in the ARIC white population.

The QT interval reflects the time between the initial fast depolarization of the ventricle and its repolarization. Several mechanisms have been reported to lead to QT prolongation: autonomic system imbalance and autonomic neuropathy (27), mutations of genes affecting cardiac ion channels involved in cardiac repolarization (28), nonconducting scar tissue resulting from myocardial infarction (1), high glucose level (29- 30), elevated insulin level (30- 32), hypokalemia (32), obesity (33), and ventricular hypertrophy (16). All these mechanisms seemed to contribute to QT prolongation in the ARIC population. The QTc prolongation was associated with the presence of ECG abnormalities, possibly resulting from small silent myocardial infarctions; QTc duration was also strongly related to all cardiovascular risk factors, including fasting insulin. This has previously been observed in Dutch elderly men (6) and in the nondiabetic subjects of the Insulin Resistance and Atherosclerosis Study (IRAS) (31). The presence of all these factors of the insulin resistance syndrome is expected to lead to atherosclerosis, and, indeed, we confirmed the association between intima-media thickness and QTc, also reported in IRAS. Thus, the QTc may be viewed as a marker of subclinical atherosclerosis, which indeed may contribute to the increased incidence of symptomatic coronary heart disease associated with QTc prolongation. However, the simultaneous presence of other cardiovascular risk factors, ECG abnormalities, and intima-media thickness did not completely explain the association between QTc prolongation and either CHD incidence or CVD mortality. Gastadelli et al. (32) recently reported a direct effect of exogenous insulin on QTc, which was mediated through lowering of serum potassium and an increase of noradrenaline levels. It was suggested that insulin-stimulated cellular potassium uptake leads to hyperpolarization of the cell membrane. In the present study, adjustment for fasting insulin or potassium level also did not explain the increased risk of QTc prolongation. But this may be only one of several possible mechanisms of (acquired) suboptimal regulation of cardiac ion channels involved in ventricular repolarization. In analogy with the findings in subjects with the long QT syndrome (28), this may directly be related to susceptibility to both fatal and nonfatal arrhythmias.

Some experimental work suggests subjects with QTc prolongation may benefit from preventive treatment. In obese subjects who lost weight, QT-interval duration normalized (33). In addition, hypertensive rats treated with angiotensin-converting enzyme inhibitors showed a reduction of left ventricular mass and normalized QT (34).

In conclusion, apparently healthy men and women with QT prolongation are at increased risk of incident CHD, and especially of CVD mortality. Therefore, QTc is a cheap and noninvasive tool to detect high-risk subjects who should receive special attention.