OUP user menu

QTc interval and survival in 75-year-old men and women from the general population

Göran Nilsson, P. Hedberg, T. Jonasson, I. Lönnberg, J. Öhrvik
DOI: http://dx.doi.org/10.1093/europace/euj040 233-240 First published online: 13 February 2006


Aims The study concerns the relationship of the corrected QT (QTc) interval to 6.4 years of survival and to measures of cardiac function, such as echocardiographic variables and plasma levels of brain natriuretic peptide (BNP), in 75-year-old people.

Methods and results QTc was measured in a 12-lead electrocardiogram (ECG) in 210 men and 223 women, comprising a randomly selected sample from the general population (70% participation rate). The Sicard 440/740 computer-analysis program, with Hodges' formula for heart rate-based QT correction, was used. The optimal cut-off point for predicting survival according to the receiver operating characteristic curve was found between 429 and 430 ms. Individuals with a QTc interval of ≥430 ms (n=115) had decreased survival when compared with those with shorter QTc interval (n=318); the relative risk was 2.4 (95% confidence interval 1.5–3.7). The predictive ability of QTc reflects an association between QTc and the following variables: BNP, left ventricular mass, and left ventricular ejection fraction (but not diastolic filling patterns). Both Hodges' and Bazett's formulae for heart rate correction of the QT interval were useful for predicting survival. The median QTc was 415 ms using Hodges' formula and 430 ms with Bazett's formula. The QRS component of QTc predicted survival better than the rest of the QTc interval and was approximately as useful as the QTc interval itself.

Conclusion The computer-derived QTc obtained from the ordinary 12-lead ECG identifies high-risk individuals among elderly people from the general population.

  • ECG
  • QT
  • BNP
  • Echocardiography
  • Risk stratification
  • Heart rate


Most, but not all, studies suggest that a prolonged QT interval corrected (QTc) for heart rate is associated with an increased risk of both cardiovascular and all-cause mortality in the general population.14 The increase in mortality appears to be restricted to subsets with signs of cardiovascular disease, suggesting that a relatively long QTc may be a marker of cardiovascular disease. A relationship has also been found between QTc and carotid atherosclerosis in non-diabetic patients, suggesting that a long QTc interval may be a marker of subclinical atherosclerotic disease.5 In addition, a long QTc interval may be associated with dysfunction of the autonomic nervous system.6

QTc increases with age.7,8 Population-based studies of the survival implications of the length of the QTc interval usually recruit people aged <65, but in the Rotterdam study9 a cohort of elderly subjects (a few >75 years of age) were followed for 3–6 years. The increase in all-cause mortality associated with the prolongation of QTc was less pronounced in people >70 years of age than in younger age groups.

There is a need to extend our knowledge of the relationship between QTc length and survival and the extent to which the QTc interval reflects cardiac function in elderly people.


The primary objective of the present study was to determine the relationship of all-cause mortality, as well as of cardiovascular death, with the length of the QTc interval in 75-year-old men and women from the general population. It should be noted that all-cause-mortality is an objective and unbiased endpoint, unlike cardiovascular death, which may be significantly affected by ascertainment bias.10

A secondary objective was to examine whether the predictive ability of QTc with regard to survival was related to established signs of reduced cardiac function, such as echocardiographically determined systolic and diastolic left ventricular function, plasma concentration of B-type or brain natriuretic peptide (BNP), and signs of a pathological ECG as determined by the Minnesota code. Other secondary objectives were to examine the impact on survival prediction of different formulae for heart rate correction and to define the relative importance for survival of the QRS and the JTc (also called non-QRS) components of the QTc interval.


Study population

The details of the Västerås study of 75-year-old men and women have previously been published.11 In short, 433 persons (210 men and 223 women) born in 1922, representing 70% of a random sample of 618 men and women born in 1922 and living in the town of Västerås, Sweden, were extensively examined in 1997 in terms of their cardiovascular health. The reasons for non-participation by 185 persons were that the person could not be reached (n=29), the person died before the investigation procedure was initiated (n=2), language difficulties and logistical problems (n=26), locomotor impairment (n=28), treatment for heart disease (n=13), other diseases (n=41), and unknown (n=46). None of the participants was using drugs associated with the prolongation of the QT interval.

The survival status was determined on 1 September 2003 (median follow-up 6.4 years; maximum follow-up 6.6 years). Death certificates were obtained from the Epidemiological Centre, Swedish National Board of Health and Welfare. Cardiovascular death was defined as ICD codes I21 through I71. The research Ethics Committee at the University of Uppsala, Sweden, approved the study. The study complies with the Declaration of Helsinki.

Self-reported previous myocardial infarction (confirmed from medical records), angina pectoris, hypertension, and diabetes were registered. Beta-blockers and calcium inhibitors (Table 1) were the only antiarrhythmic drugs used in the patients of present study. No patient was taking sotalol.

View this table:
Table 1

Baseline characteristics of the study patients according to survival

Clinical characteristicsSurvivors (n=353)Non-survivors (n=81)P-value
Male gender44.569.1<0.001
History of myocardial infarction7.717.30.012
Angina pectoris11.122.20.011
Current smokers8.219.80.019
Calcium inhibitors13.06.10.088
Lipid-lowering drugs2.86.20.171
Antidepressant drugs1.12.50.313
Systolic blood pressure165.5±25.0163.1±25.30.457
Diastolic blood pressure84.3±9.884.3±10.50.998
Echocardiographic variables
 Wall motion index2.04±0.261.90±0.41<0.001
 EF (Simpson)60.0±9.053.6±13.9<0.001
 Left ventricular mass87.7 (75.3–104.4)95.5 (84.7–123.7)0.002
 BNP (pg/mL)25.4 (12.5–48.0)35.1 (19.5–81.4)0.001
Ordinary 12-lead ECG
 Normal according to the Minnesota code67.450.60.007
 Abnormal Q-wave4.09.50.087
 Bundle branch block4.55.20.775
 Atrial fibrillation4.28.60.154
 QTc (ms)413 (401–428)425 (407–439)0.002
 QRS (ms)90 (82–98)94 (85–110)0.002
 JTc (ms)322 (310–337)324 (311–342)0.331
  • Values are presented as the mean±SD, median (interquartile range), or percentage.

  • BMI, body-mass index; ACE, angiotensin converting enzyme; EF, ejection fraction.


A standard 12-lead electrocardiogram (ECG) was taken using a Siemens Elema AB (Solna, Sweden) machine. The QT and QRS intervals were measured by the Sicard 440/740 ECG computer-analysis programme (Megacart version 3 V4, 7/2.38/23, Siemens Elema),12 which is used worldwide and has been extensively validated.13

In this programme, an overall onset and termination were computed for QT and QRS from the average beat. The QT intervals were corrected for heart rate using Hodges' linear-correction formula [QTc=QT+1.75(heart rate−60)].14 This formula was chosen because the QT interval corrected with this formula was found to have a negligible correlation with heart rate as opposed to the extensively used Bazett's formula.15 However, in the present study, survival was also determined after correcting the computer-derived QT interval according to the formulae of Bazett15 (QTc=QT/square root of RR interval), Fridericia16 (QTc=QT/cubic root of RR interval), and Rautaharju.17

The heart rate-corrected JT interval, JTc, was defined as QTc minus QRS.18

The ECG was registered in the morning after a resting period of at least 10 min. Two physicians blinded to the clinical data of the patient coded the ECG according to the Minnesota system.19 If the coding differed between the physicians, a new coding was performed by consensus. Using the Minnesota code, the ECG was classified as normal in the absence of the following major abnormalities20: abnormal Q-wave, ST-segment depression or elevation, T-wave change, incomplete or complete left or right bundle branch block, atrial fibrillation, atrioventricular block, left-axis deviation, and high R-wave amplitude.


Echocardiography was performed using an Acuson XP 128 system (Acuson Co, Mountain View, CA, USA). The same physician (P.H.), who was blinded to the participants' clinical data, performed all the studies. Left ventricular ejection fraction (LVEF) was calculated online using the biplane disc summation method (modified Simpson's rule) in participants in whom at least 60% of the endocardial border could be detected (n=279). A wall motion index for the left ventricle was computed by dividing the left ventricular wall into nine segments examined in five standard projections, as described in detail by Hedberg et al.11 This measure was available in 411 cases. In addition, adjusted left ventricular mass was calculated as left ventricular mass/body surface area in square metres, which was calculated according to the formula of Dubois [length (in cm)0.425*weight (in kg)0.725*71.84].

Using the Doppler technique, the following measures representing diastolic filling patterns of the left ventricular chamber were obtained: peak atrial (A-wave) and early diastolic (E-wave) velocities for transmitral flow, quotient between A- and E-waves, and deceleration time of the E-wave. Some measurements were not available in all patients (Table 2) for various reasons, such as the lack of the A-wave in the cases of atrial fibrillation.

View this table:
Table 2

Characteristics of patients with QTc <430 ms and patients with QTc of 430 ms or longer

QTc<430 ms (n=318)QTc≥430 ms (n=115)P-value
Previous myocardial infarction7.215.80.008
Angina pectoris9.124.6<0.001
Diabetes mellitus8.27.0ns
Current smokers11.66.9ns
Atrial fibrillation4.75.6ns
Abnormal ECG26.760.3<0.001
Dead after 6.4 years11.027.6<0.001
BNP (pg/mL)23.0 (11.8–45.2)37.9 (19.1–79.7)<0.001
Systolic blood pressure164.0±25.1167.9±24.7ns
Diastolic blood pressure83.7±9.785.8±10.2ns
Electrocardiographic variables
 QRS interval (ms)88 (82–96)96 (86–114)<0.001
 JTc interval (ms)317±16346±23<0.001
Echocardiographic variables
 Left ventricular mass (g/m2, body surface), n=245/9487.4 (76.4–101.5)95.8 (78.9–125.0)<0.001
 Left ventricular wall motion index, n=302/1092.1±0.31.9±0.4<0.001
 EF, n=204/7660.3±8.954.3±13.0<0.001
 Peak velocity E-wave, n=305/1070.57 (0.49–0.67)0.61 (0.50–0.72)0.047
 Peak velocity A-wave, n=290/1010.68 (0.54–0.80)0.69 (0.57–0.80)ns
 E/A ratio, n=290/1010.84 (0.69–1.03)0.84 (0.72–1.05)ns
 E-wave deceleration time (ms), n=298/104224 (195–262)232 (198–270)ns
  • Values are presented as the mean±SD, median (interquartile range), or percentage. Abbreviations as in Table 1.

Laboratory methods

Venous blood was sampled in the morning from participants who had been resting in a recumbent position for at least 5 min. For the analysis of BNP, the blood was collected in 10 mL ice-chilled tubes containing ethylenediamine tetraacetic acid. The tubes were turned 5–10 times and placed again on ice, centrifuged at 4°C at 200 g. The separated plasma was then put into polypropylene tubes and frozen at −70°C. BNP analyses were performed at the Western Infirmary, Glasgow, UK, using two-site monoclonal antibody immunoradiometric assays (Shionoria BNP kit, Shionogi & Co. Ltd, Osaka, Japan). The within-assay and between-assay coefficients of variation were 3.7 and 7.5%, respectively. The concentration is presented in pg/mL.

The concentration of total cholesterol, triglycerides, and high-density cholesterol (HDL) were measured using routine methods; low-density cholesterol (LDL) was calculated using Friedewald's formula.

Blood pressure was measured to the nearest 5 mmHg with a mercury sphygmomanometer with subjects sitting and relaxed for 10 min.


The Wilcoxon Mann–Whitney rank-sum test and Student's t-test were used to test differences between groups containing data with non-normal and normal distribution, respectively. Categorical data were compared using χ2 statistics.

Receiver operating characteristic (ROC) curves were calculated to determine an optimal cut-off point, which was defined as the QTc length resulting in the highest sum of the sensitivity and specificity.21 The purpose of this calculation was to define a QTc level suitable as a rule of thumb for finding individuals with a high mortality risk in clinical practice. Ten-fold cross-validation was performed to compensate for the fact that the same individuals were used to assess the optimal QTc cut-off level and for the classification into survivors and non-survivors.22

Cumulative mortality was estimated by means of the Kaplan–Meier analysis comparing participants below and above the optimal QTc interval cut-off level found by ROC curve analysis. Difference in survival was calculated according to the log-rank statistics.

We calculated QTc group differences, hazard ratios, and 95% confidence intervals (CIs) with the Cox-proportional hazard regression model, in both univariable and multivariable analyses. Hazard ratios were used to estimate relative risks. A forward stepwise analysis using 0.05 as entry probability and 0.10 as removal probability was used in the adjusted analysis. Missing values for blood pressure (n=11), LDL cholesterol (n=9), HDL cholesterol (n=1), and BNP (n=1) were replaced by the mean value for individuals with available measurements in the adjusted analysis.

Pearson's product moment correlation coefficient was used to assess the association between variables.

A two-sided P-value of less than 0.05 was considered to be statistically significant. SPSS version 11.0 was used for statistical analysis.


In all, 81 persons died during the 6.4 year follow-up. The number of people who died of cardiovascular disease was 36 (ischaemic heart disease 12, other types of heart disease 12, non-cardiac atherosclerotic disease 12). Among the group with non-cardiac atherosclerotic diseases, seven cases died of cerebrovascular disease.

The essential basal clinical characteristics of the patient cohort are shown in Table 1.

The distribution of the QTc in the present cohort is shown in Figure 1. The median (interquartile range) of QTc was 415 (401–431) ms for all the participants in the study, 413 (401–428) ms for survivors, 425 (407–439) ms for non-survivors, 433 (416–450) ms for patients dying of cardiovascular disease, and 419 (402–436) ms for patients dying of non-cardiovascular disease.

Figure 1

Distribution of QTc interval length in 433 people, aged 75, comprising a randomly selected sample from the general population.

The area under the ROC curve, determining the sensitivity and specificity of different lengths of QT interval to predict survival, decreased steadily with time and was 0.716 after 1 year, 0.674 after 2 years, 0.647 after 5 years, and 0.614 (95% CI 0.543–0.684) after 6.4 years. At all these points in time, this implies a significant deviation at the 0.001 level from the null hypothesis that the true area under the ROC curve is 0.5. The optimal cut-off point (maximum sum of sensitivity and specificity) of QTc interval length for predicting survival was found between 429 and 430 ms, at both 5 and 6.4 years. All the QTc intervals were measured to the nearest whole number (in ms), implying that this classification corresponded to a dichotomization into <430 ms/≥430 ms. Using 10-fold cross-validation, a specificity of 78%, a sensitivity of 41%, and a total misclassification rate of 29% were found at this QTc length.

A Kaplan–Meier curve for individuals with a QTc interval above and below the cut-off point is shown in Figure 2. The log-rank statistic was 15.57, one degree of freedom (P<0.001). As it is well known23 that left bundle branch block, as opposed to right bundle branch block, is associated with a dismal survival, we repeated the analysis after exclusion of cases with left bundle branch block (n=5). The log-rank statistic then was 15.02 (P<0.001).

Figure 2

Kaplan–Meier cumulative probability of survival by QTc length dichotomized at 430 ms. Upper line ≥430 ms (n=115) and lower line <430 ms (n=318).

Restricting the analysis to cases with cardiovascular death resulted in a Kaplan–Meier curve similar to that with total mortality; log-rank statistic was 18.58 (P<0.001). The corresponding P-value for non-cardiovascular deaths was 0.051, with a better survival for patients with QTc intervals below 430 ms. There was no difference in survival between the first and second tertiles of QTc. The third tertile (QTc>424 ms) differed significantly (log-rank statistic 10.30; P=0.001) from the first tertile by the Kaplan–Meier analysis.

The relative risk was 2.4 (95% CI 1.5–3.7; P<0.001) for persons with a QTc≥430 ms by comparison with persons with shorter QTc interval.

Combination of QTc and BNP determinations

Serum BNP is often determined in people with suspected cardiac disease. Determination of the optimal cut-off point for BNP in the same way as the determination of the cut-off point for QTc resulted in a cut-off point of 73 pg/mL (sensitivity 29%, specificity 89%). BNP >73 pg/mL in combination with QTc>430 ms delineated a group with 56% 6 year survival when compared with a 90% survival rate for patients with QTc<430 ms in combination with BNP <73 pg/mL. The first group comprised 7% and the latter group 68% of the participants in the study. Log-rank statistic by the Kaplan–Meier analysis was 36.82 (P<0.001).

Characteristics of participants with QTc above and below 430 ms

In Table 2, some data, including echocardiographic, are shown for patients with QTc <430 ms and for those with QTc of 430 ms or longer.

QTc and conventional risk factors

Univariable Cox-regression analysis of relative risks of death associated with an increase of 1 SD in various risk factors (Table 3) was performed in order to put the relationship of QTc with survival into perspective. The length of the QTc classified by the optimal cut-off between 429 and 430 ms interval performed very well in predicting survival, in comparison with conventional risk factors.

View this table:
Table 3

Association between various risk factors and survival according to the univariable Cox-regression analysis

Relative risk (95% CI)P-value
Conventional risk factors
 Systolic blood pressure0.92 (0.73–1.17)ns
 Diastolic blood pressure0.93 (0.80–1.27)ns
 LDL cholesterol0.77 (0.62–0.97)0.027
 HDL cholesterol0.62 (0.46–0.83)0.001
 Triglycerides1.07 (0.88–1.30)ns
 BMI0.88 (0.70–1.11)ns
Cardiac risk factors
 BNP1.30 (1.19–1.42)<0.001
 Left ventricular wall motion index0.70 (0.59–0.82)<0.001
 LVEF0.61 (0.50–0.76)<0.001
 Left ventricular mass (m2)1.42 (1.19–1.69)<0.001
 QTc1.40 (1.15–1.71)0.001
  • Relative risk associated with an increase of 1 SD in the individual variable.

Forward stepwise Cox-regression analysis, including, in addition to QTc, those conventional risk factors from Table 3 that were significantly related to survival, yielded the following significant variables in the final model: QTc (P=0.004), HDL cholesterol (P=0.001), and LDL cholesterol (P=0.018). Relatively low serum levels of both HDL and LDL cholesterol predicted dismal survival. Adding BNP to the model resulted in the following variables as significant risk factors: QTc (P<0.001), BNP (P<0.001), HDL (P=0.003), and LDL (P=0.024).

QTc and echocardiographically determined indices of left ventricular function

As indicated in Table 2, the length of QTc is closely associated with echocardiographically determined left ventricular wall-motion index, LVEF, and left ventricular mass adjusted for body surface. Including these variables, together with the significant variables (QTc interval, BNP, HDL, and LDL) in a forward stepwise Cox-regression analysis, yielded LVEF (P=0.002) and BNP (P=0.024) as significant predictors of survival. It should be noted, however, that this model included only 280 of 433 patients, because LVEF could only be determined if at least 60% of the endocardial border could be visualized.

Survival related to different heart rate correction formulae for length of the QT interval

Hodges' linear-correction formula [QTc=QT+1.75(heart rate−60)]14 is the heart rate correction in the Sicard analysis programme we used. The extensively used Bazett's formula (QTc=QT/square root of RR)15 resulted in a median value of 430 ms when compared with 415 ms using Hodges' formula.

Both QT intervals uncorrected for heart rate and QT intervals corrected for heart rate by Hodges', Bazett's, Fridericia's, and Rautaharju's formulae significantly predicted mortality (Table 4). Hodges' formula showed the highest value of relative risk.

View this table:
Table 4

Association between 6.4 year survival and QT corrected for heart rate by different formulae as well as components of QTc

Relative risk (95%CI)P-value
QT1.31 (1.05–1.64)0.015
Heart rate correction formulae
 QTc (Hodges)1.40 (1.15–1.71)0.001
 QTc (Bazett)1.29 (1.06–1.58)0.013
 QTc (Fridericia)1.34 (1.10–1.65)0.004
 QTc (Rautaharju)1.34 (1.10–1.64)0.004
Components of QTc (Hodges)
 QRS1.35 (1.14–1.61)0.001
 JTc1.12 (0.91–1.39)ns
  • Relative risk associated with an increase of 1 SD in the variable.

Survival related to QRS and non-QRS (also called JTc) components of the QTc interval

The QTc interval consists of two components: first the QRS interval and then the non-QRS interval, usually known as the JTc interval. These two components of the QTc interval were analysed separately with regard to 6.4 year survival (Table 4).

This table indicates that the QRS component of the QTc interval is much more important for survival than the JTc component of the QTc interval. The QTc interval was positively correlated with both QRS interval (P<0.001) and JTc interval (P<0.001), with Pearson's correlation coefficients 0.417 and 0.768, respectively. The QRS interval was inversely correlated with the JT interval (r=−0.261; P<0.001). A forward stepwise Cox-regression analysis, including QRS and JTc, resulted in a model including both these intervals, but with a higher significance for QRS (P<0.001) than for JTc (P=0.031). In the present investigation, the QRS component of the QTc interval is thus more closely related to survival than the JTc component of the QTc interval.

It is noteworthy that the ability of the QRS interval to predict survival is approximately the same as that of the QTc interval (Table 4). A forward stepwise Cox-regression including these two intervals yielded practically the same significance for QRS (P=0.035) as for QTc (P=0.029). No relationship between heart rate and the QRS interval was found.


The QT interval is an indirect measure of the ventricular action potential, including the depolarization and repolarization of the ventricular chambers. There is a physiological relationship between increasing length of the QT interval and increasing heart rate. This has led to the convention of correcting the QT interval for heart rate by means of various formulae prior to clinical evaluation of the interval. There are some inherent problems when measuring the QT interval on an ordinary 12-lead surface ECG, mainly because of the difficulty in determining the end of the T-wave and evaluating the U-waves.

Computerized QTc interval measurement

Several investigators have used computerized measurement of the QTc interval. The computerized measurement in the present investigation was developed by Macfarlane et al.12 and is used in the Sicard system, which is an extensively used computer-analysis programme. It was originally developed for the Siemens Elema ECG machine. At our department, the QTc intervals are shown on all paper printouts of ordinary 12-lead surface ECGs. A thorough knowledge of the clinical importance of the varying length of the QT interval is therefore of great value.

The present investigation demonstrates for the first time the predictive power for survival of the QTc algorithm in the Sicard computer-analysis programme.

QTc interval and survival

The adverse prognostic importance of a relatively long QTc interval has previously been demonstrated in the majority of population-based studies.1,2,4,24 Conflicting results could reflect imprecision in the measurement of the QTc interval, a varying prevalence of cardiac disease at baseline and different lengths of follow-up. It is essential to note that the QTc interval increases with age.25 Our findings corroborate and extend previous work9 on the association between the length of the QTc interval and survival in elderly people from the general population. Studies of such a homogeneous age group of elderly people from the general population have not previously been reported.

The prognostic importance of the QTc interval probably reflects structural abnormalities in the ventricular heart muscle, most importantly reduced left ventricular systolic function and increased left ventricular mass, which were closely related to the length of the QT interval. The prognostic ability was practically unchanged after exclusion of persons with left bundle branch block, which is known to be associated with poor survival.23 Furthermore, the QTc interval is determined at least in part by the activity of the autonomic nervous system,26 and a relatively long QT interval has been associated with sudden cardiac death among diabetic patients with abnormal autonomic function.27

In a clinical context, a QTc interval of ≥430 ms should suggest further investigation with BNP and/or echocardiography. It must be noted that this figure is calculated by Hodges' formula and corresponds to 415 ms with Bazett's formula. In patients with advanced heart failure, Vrtovec et al.28 have observed a prognostic value of the combination of QTc and BNP in excess of the value of QTc alone. This is analogous to our observations in a general population.

In the present investigation, QTc has a much closer relationship with systolic than with diastolic left ventricular function. To our knowledge, data on the relationship between QTc and diastolic function have not previously been reported.

QTc and conventional risk factors

Information on the length of the computerized QTc interval is easy to obtain and performs very well as a prognostic instrument when compared with established risk factors such as blood lipids and blood pressure. Systolic and diastolic blood pressure, as well as triglycerides and body-mass index failed to show a significant relationship with survival, indicating that these traditional risk factors have a reduced ability to predict prognosis in elderly people. In the present study, the strongest conventional risk factor in this age group was low HDL cholesterol; surprisingly, low LDL cholesterol is also a significant risk factor for death in this age group and is not the good prognostic factor that it is in younger people. An ECG with measurements of QTc interval is therefore a simple and inexpensive method for detecting high-risk individuals among the elderly. In addition, the presence of a normal ECG according to the Minnesota code is highly useful to predict left ventricular systolic function as demonstrated by our group.20

Heart rate correction formulae

There have been many discussions about the best formula for heart rate correction of the QT interval. The correction method that best correlates with clinical outcome and survival is advantageous from a clinical point of view. Data that relate clinical outcome to QT interval corrected by means of different formulae have previously been reported by de Bruyne et al.9 The formulae tested by them included the extensively used Bazett's formula but not the linear Hodges' formula, which was used in the present investigation. They found that survival was hardly affected by the heart rate correction formulae tested.

Our investigation indicates that the Hodges' formula,14 which is routinely used in the Sicard 440/740 ECG computer-analysis programme, performs well in comparison with Bazett's formula,15 which probably is the most extensively used formula for heart rate correction. The relationship between survival and QTc intervals, computed with Hodges' formula, has not previously been reported. However, it must be noted that the relationship between heart rate and the length of the QTc interval shows extremely high interindividual variation.29

Relative importance of the QRS and JTc components of the QTc interval

QTc reflects the duration of both the QRS interval and the JTc interval. In routine ECGs from a clinical institution, Banker et al.30 described a reciprocal relationship between QRS and JTc in a general population, similar to that observed by us.

The greater predictive value for survival of the QRS component of the QTc interval when compared with that of the JTc component of the QTc interval is noteworthy and, to our knowledge, it has not previously been reported for a general population cohort. In actual fact, the QRS interval itself significantly predicts survival. QRS width has previously been reported to be a prognostic factor in chronic heart failure.31 In men with a QRS interval >120 ms, Crow et al.18 demonstrated a statistically significant predictive value also of JTc for 13 year incident cardiovascular events in a population-based prospective cohort.

The prognostic value of the QRS interval is of obvious clinical relevance. The QRS interval is more easily defined than the QTc interval, and as opposed to the QTc interval, the QRS complex is unaffected by heart rate.

Strengths and limitations of the study

The present investigation has some important strengths. Theuniform age of the participants prevented confounding by age. Holter studies have demonstrated a diurnal variation in the QTc interval32 and our recording of the ECG at the same time of the day (i.e. in the morning after at least 10 min rest), thus facilitates the comparison of QTc interval between different persons. People as old as 75 have previously only been studied to a very limited extent with regard to the prognostic importance of the length of the QTc interval. Our cohort is reasonably representative of the general population, because as many as 70% of the inhabitants in a restricted area participated in the study. The investigation also offers unique data on the relationship between QTc interval and diastolic filling patterns, as well as on the relationship in elderly people between survival and the QTc corrected by varying formulae. Furthermore, data on the relative importance of the QRS and JTc components of the QTc interval have previously not been reported.

A limitation in the present investigation is the relatively small cohort, as well as the missing data, especially on ejection fraction determined echocardiographically according to Simpson. This is, however, unavoidable because of the difficulty in defining the endocardial borders in many persons. Furthermore, the present study does not evaluate the possibility that the relationship of QTc to survival reflects the function of the autonomic nervous system.


In conclusion, the present investigation shows that the length of the QTc interval is a valuable predictor of survival in elderly men and women in the general population. The computer-derived QTc is a simple and inexpensive method of detecting high-risk individuals requiring special medical attention in the general population.


This work was supported by grants from Västmanland's research foundation against cardiovascular disease and Sparbanksstiftelsen Nya.


View Abstract