In the Framingham Heart Study, we report findings that are consistent with multiple other reports from other cohorts on the association of the continuous measures of QTC adjusted for heart rate using Bazett’s correction with all-cause mortality, CHD mortality, and SCD in models adjusting for age and sex. However, we demonstrate that additional adjustment for RR interval, either by additional adjustment of QTC for RR interval or by use of QT interval adjusted for RR interval by linear regression, substantially attenuates the association. Additional adjustment for baseline clinical characteristics further attenuates the association of QTC with all-cause or CHD mortality. These data suggest that (1) the Bazett’s formula incompletely adjusts for the heart rate association with mortality, (2) some of the association of QTC with mortality is accounted for by other clinical factors, and (3) that the QT interval itself is a modest contributor to CHD mortality and SCD risk beyond other routinely available clinical data.
Our results differ from some prior reports, although direct comparisons between studies is challenging because results have been reported using different QTC
definitions/cut-points, outcomes, and populations. Strauss and colleagues examined subjects in the Rotterdam study (a prospective population-based cohort with nearly 8,000 individuals and 125 adjudicated SCD cases) and showed that a prolonged QTC
(>450 msec in men and 470 msec in women, using Bazzet correction23
) was associated with more than three-fold risk of SCD in a model adjusting for age alone (HR 3.7, 95% CI 2.0–6.9). This association was somewhat attenuated, but remained significant, after additional adjustment for CHD risk factors and heart rate (HR 2.5, 95%CI 1.3–4.7).6
Our study is of comparable size (in both controls and SCD cases) so the reason for this discrepant finding is not likely to result from power alone. One difference between the studies is that ECG data from participants in the Rotterdam study who had a second ECG during a follow-up visit were included. Thus, if QT prolongation developed after several years of follow up, and reflected a change in the underlying cardiac substrate, it could contribute to the observed association. This approach answers a different question from the prognostic implications of the baseline QT interval for lifetime SCD risk, and, instead, could reflect the dynamic nature of the ECG as cardiac risk evolves though a patient’s lifetime. Indeed, an earlier examination of the association in the Rotterdam study showed a more modest risk increase (HR 1.7, 95% CI 1.0–2.7) for cardiac mortality associated with the top quartile (different QTC
cut points were used in these two studies) of QT interval when only the baseline ECG was used.24
The relations of QT interval to cardiovascular mortality has been examined in several other epidemiologic studies, but the results have been inconsistent.5–14, 25, 26
For instance, the Cardiovascular Health Study, showed a roughly two fold risk of all-cause mortality (attenuated somewhat after adjustment for cardiac risk factors, measures of atherosclerosis, and heart rate) above a QTC
of 450 msec,13
the Zutphen study showed a three-fold increased risk of SCD with QTC
prolongation above 420 msec (HR, 3.0; 95% CI, 1.0 to 8.9) in men ages 65 to 85 years, but not in younger men,8
and the first report in the Framingham Heart Study showed no association between baseline QTC
prolongation and all-cause mortality, sudden death, or coronary mortality.14
More recently, a large study from the Third National Health and Nutrition Examination Survey demonstaated increased risk for mortality due to cardiovascular disease (HR 2.55, 95% CI 1.59–4.09) comparing the 95th percentile of age-, sex-, race-, and RR interval–corrected QT interval with participants in the middle quintile. Additionally, they showed that there appeared to be a U-shaped relationship between QT interval and mortality (i.e. higher mortality at with a shortened or prolonged QT-interval). Our study did not demonstrate this U-shaped relationship.27
Differences in these estimates could be explained by differences in the method of QT measurement or cut-point definition, differences in endpoint definition and adjudication, differences in population characteristics, or duration and method of follow-up.
We found the unadjusted risk of all-cause mortality increased in a graded fashion across the range of QTC
quintiles. Although non-significant, this trend persisted after adjustment for additional risk factors and was observed for CHD mortality and SCD. This observation supports the concept that the QT interval may reflect incremental risk across the range of “normal” values, rather than only above a particular threshold of an extreme QT. Furthermore, it provides some motivation for the study of the recently discovered QT-modifying genetic variants in relation to SCD,28, 29
but since the QT-SCD association is modest, the effect of these variants on mortality may be small individually.
The observation of residual confounding by heart rate, even after Bazett’s correction, is not surprising. Heart rate and QT interval are inextricably linked. Heart rate is the principal determinant of the QT interval,30
and is, itself, a predictor of SCD and all-cause mortality.31, 32
Furthermore, separating the determinants of heart rate, QT interval, and SCD risk is challenging because heart rate is heritable,33
and by some estimates, 40% of the heritability of the QT interval could be through genes that also contribute to heart rate.34
The major cardiovascular societies as well as the Food and Drug Administration have made formal recommendation against the use of the Bazett formula in research, drug development and testing, and clinical practice. Indeed, the AHA/ACCF/HRS Recommendations for the Standardization and Interpretation of the Electrocardiogram advise against the use of Bazetts formula and expressly encourage use of linear regression functions.35
However, the Bazett formula remains the most commonly used correction in routine clinical practice, likely due habit, ease of use, and the relative difficulty of applying alternate correction methods (especially population-specific regression-based methods).
We show that additional correction for clinical factors known to increase CHD and SCD risk (namely total cholesterol, HDL, smoking status, SBP, DBP, antihypertensive medication use, BMI, and DM) attenuates the association between QT interval and mortality. Thus, clinical factors explain a portion of the QT-associated mortality risk and could be considered with QT in clinical SCD risk assessment. Although there is great interest in refining SCD risk stratification, for indications for implantable cardioverter defibrillator (ICD) implantation for example, it is unlikely that considering a single baseline QT interval will add substantially to current strategies.
The strengths of the current study include its large sample of prospectively followed individuals and community-based cohort ascertainment, detailed clinical information, rigorous outcome adjudication, reproducible ECG measurements, and long duration and minimal loss to follow up. The major limitation is the study’s low power to show association with SCD given relatively few cases.