Prolongation of QTc interval and risk of stroke: the REasons for Geographic and Racial Differences in Stroke (REGARDS)

Abstract

Objective

To examine the association between prolongation of heart rate–corrected QT interval (QTc) with incident stroke.

Background

Unlike cardiovascular morbidity and mortality, little is known about the relationship between QTc and risk of stroke.

Methods

A total of 27,411 participants aged ≥ 45 years without prior stroke from the REasons for Geographic and Racial Differences in Stroke (REGARDS) study were included in this analysis. QTc was calculated using Framingham formula (QTc_Fram). Stroke cases were identified and adjudicated during up to 8.2 years of follow-up (median 5.1 years).

Results

The risk of incident stroke in study participants with prolonged QTc_Fram was almost three times the risk in those with normal QTc_Fram [HR (95% CI): 2.88 (2.12, 3.92), p<0.0001]. After adjustment for demographics (age, race, sex), traditional stroke risk factors (antihypertensive medication use, systolic blood pressure, current smoking, diabetes, left ventricular hypertrophy, atrial fibrillation, prior cardiovascular disease), warfarin use, aspirin use, QRS duration and use of QT-prolonging drugs, the risk of stroke remained significantly high [HR (95% CI): 1.67 (1.16, 2.41), p=0.0061)], and was consistent across several subgroups of REGARDS participants. Similar results were obtained when the risk of stroke was estimated per 1-standard deviation increase in QTc_Fram, [HR (95% CI): 1.12 (1.03, 1.21), p=0.0053 in multivariable-adjusted model], and when other QTc correction formulas including Hodge’s, Bazett’s and Fridericia’s were used.

Conclusions

QTc prolongation is associated with a significantly increased risk of incident stroke independent of traditional stroke risk factors. Examining the risk of stroke associated with QT-prolonging drugs may be warranted.

INTRODUCTION

The association between heart-rate corrected QT (QTc) interval and cardiovascular morbidity and mortality is well established (1–4). Little is known, however, about the relationship between this simple electrocardiographic (ECG) marker and incident stroke.

Prolonged QTc interval has been reported in 38–71% of patients during acute stroke, constituting the most frequent single ECG abnormality in this setting (5). The pathophysiology behind this phenomenon has not been elucidated. Autonomic disregulation caused by overactivity of the sympathetic nervous system during acute stroke has been suggested as one of the potential mechanisms (6). Nevertheless, it is also possible that in some cases prolonged QTc actually existed before developing stroke and its presence during the acute phase reflects its common presence in individuals at risk; hence it could be used as a marker for future stroke. Further, prolonged QTc is associated with several cardiovascular disease (CVD) risk factors including advanced age, impaired glucose homeostasis, smoking, left ventricular hypertrophy (LVH), and high blood pressure (7, 8), all of which are known risk factors for stroke. Therefore, it is plausible that QTc has prognostic value for prediction of future stroke. This contention is supported by a small study from Brazil which included 471 individuals with diabetes (9) and in a group from a general Japanese population (10), but no data exists from the US or European population-based studies.

We sought to examine the risk of incident stroke associated with prolongation of QTc interval in the REasons for Geographic and Racial Differences in Stroke (REGARDS) study. With its biracial population, centrally read ECG data, physician-adjudicated stroke events and long term follow-up, the REGARDS study provides a unique opportunity to address our research questions.

METHODS

Study Population

The goals and design of the REGARDS study have been published elsewhere (11). Briefly, the study was designed to investigate the causes of regional and racial disparities in stroke mortality, oversampling blacks and residents of the southeastern stroke belt region (North Carolina, South Carolina, Georgia, Alabama, Mississippi, Tennessee, Arkansas, and Louisiana). Individuals were recruited from a commercially available list of residents using a combination of postal and telephone contact with a 49% cooperation rate. Using a computer-assisted telephone interview, trained interviewers obtained demographic information and a cardiovascular medical history. Consent was obtained initially on the telephone and subsequently in writing during an in-person evaluation. In-home brief physical examination was conducted 3–4 weeks after the telephone interview. Participants are followed every 6 months by telephone for possible stroke outcomes.

Of the 30,239 REGARDS participants enrolled between 2003 and October 2007, we excluded 1,930 participants reporting prior stroke, 359 with no or poor quality ECG data, and 539 with no follow-up data, resulting in 27,411 (91%) participants for analysis.

The Institutional Review Boards of participating centers reviewed and approved the study methods.

QTc interval

Baseline resting ECGs were recorded during in-home visits. ECG tracings were mailed for reading and coding at a central ECG core laboratory located at the Epidemiological Cardiology Research Center (EPICARE), Wake Forest School of Medicine, Winson-Salem, NC. To calculate QTc interval from the raw QT and heart rate, we followed the recommendation of the American Heart Association (AHA), American College of Cardiology (ACC) and Heart Rhythm Society (HRS) for the Standardization and Interpretation of the Electrocardiogram (12). In this context, we used the Framingham linear regression formula [QTc_Fram= QT + 154 (1–60/heart rate)] (13). QTc_Fram values of 460 ms or longer in women and 450 ms or longer in men were considered abnormal (i.e. prolonged QTc_Fram) (12). As secondary analysis to confirm the results and to provide comparability to previous studies, we also used other traditional QTc correction methods including Hodge’s (14) [QTc_Hod= QT+1.75 (HR-60)], Bazett’s (15) [QTc_Baz= QT (heart rate/60)1/2] and Fridericia’s [QTc_Frid QT (heart rate/60)^1/3] (16).

Stroke events

Report of a possible stroke/transient ischemic attack (TIA), or a positive response to the stroke symptoms on the Questionnaire for Verifying Stroke-free Status (QVSFS) (17) resulting in hospitalization, during follow-up generated a request for retrieval of medical records that were centrally adjudicated by a panel of stroke-expert physicians. Stroke events were defined following the World Health Organization (WHO) definition as “rapidly developing clinical signs of focal, at times global, disturbance of cerebral function, lasting more than 24 hours or leading to death with no apparent cause other than that of vascular origin” (18). Events not meeting the WHO definition but characterized by symptoms lasting <24 hours with neuro-imaging consistent with acute ischemia or hemorrhage were classified as “clinical strokes”. Strokes were further classified as ischemic or hemorrhagic. This analysis included WHO-defined as well as clinical ischemic or hemorrhagic fatal and non-fatal stroke cases. As secondary analysis, we used ischemic stroke only excluding hemorrhagic strokes. Further details on stroke identification and adjudication in the REGARDS study are available elsewhere (19).

Other variables

Standardized physical measures that included height, weight, and blood pressure were collected at the in-home physical examination. Demographics (age, sex and race) were defined by self-report. Traditional stroke risk factors were selected based on the components of the Framingham Stroke Risk Score which includes antihypertensive medication use, systolic blood pressure, current smoking, diabetes, LVH, atrial fibrillation (AF), and prior CVD (20). History of stroke was defined by self report of a physician diagnosis, and prior CVD was similarly defined by self report (myocardial infarction or heart attack, coronary artery bypass surgery, coronary angioplasty, or stenting) or by ECG evidence of a prior myocardial infarction. LVH was defined using the electrocardiographic Sokolow-Lyon criteria (21). AF was defined based on ECG diagnosis and/or self-report of a previous physician diagnosis as detailed elsewhere (22). Use of antihypertensive medications, current warfarin treatment, aspirin use and use of QT-prolonging drugs were defined using an inventory of current medications that was conducted during the in-home visit; all prescription and over-the-counter medications taken in the past 2 weeks were recorded. The list of drugs that prolong QTc was obtained from http://www.azcert.org/medical-pros/drug-lists/drug-lists.cfm, a web site maintained by the University of Arizona Center for Education and Research on Therapeutics.

Statistical Analysis

Frequency distributions of all variables were first inspected to identify anomalies and outliers possibly caused by measurement artifacts. Continuous data were described by their mean and SD; and categorical data, as proportions (percentage). Differences in characteristics by QT prolongation status were assessed by χ² (for categorical variables) and unpaired t tests (for continuous variables).

Cox proportional hazards analysis was used to estimate the hazard ratios for incident stroke associated with QTc_Fram interval per 1 standard deviation (SD) increase as well as prolonged versus normal, separately, in a series of incremental models as follows: First unadjusted (Model 1), then adjusted for demographics (age, sex and race) (Model 2), then additionally adjusted for traditional stroke risk factors (antihypertensive medication use, systolic blood pressure, current smoking, diabetes, LVH, AF, and prior CVD) (Model 3), and finally additionally adjusted for QRS duration, QT-prolonging drugs, warfarin use and aspirin use (Model 4). The assumptions of the Cox proportional hazards models were examined by plotting the natural log of the cumulative hazard of stroke by the natural log of time. Multiple imputation techniques for incident stroke were employed in the analysis to reduce the potential bias introduced through either failure to retrieve medical records or from records remaining in the adjudication process at the time of analysis. Additional details of the application of multiple imputation techniques in REGARDS are provided elsewhere (23).

Secondary analyses included: 1) examining the association between QTc_Fram and incident stroke across different subgroups of the study participants stratified by age, sex, race, smoking status, hypertension, diabetes, LVH, prior CVD, AF and QRS duration, 2) examining interaction between each of sex and race, with QTc_Fram for prediction of stroke, 3) using ischemic stroke only as an outcome instead of both ischemic and hemorrhagic, and finally 4) Using QTc_Hod, QTc_Baz and QTc_Frid, separately, instead of QTc_Fram in the proportional hazards models.

RESULTS

At baseline, the average QTc_Fram interval duration was 407 (+ 23) msec, with 740 (2.7%) of the participants having prolonged QTc_Fram. Table 1 shows the characteristics of the study population, overall and stratified by presence of prolonged QTc_Fram. Compared to those with normal QTc_Fram, participants with prolonged QTc_Fram were older with fewer blacks and more males. Prior CVD, diabetes, AF, LVH and use of QT-prolonging drugs were more prevalent in the prolonged QTc_Fram group.

Table 1.

Baseline characteristics of REGARDS study participants 2003–2007

Characteristic*	All population N=27411	Prolonged QTcFram† N=740	Normal QTcFram N= 26671	p-value
Age, years	64.7(9.40)	70.1(9.65)	64.5(9.34)	<0.0001
African American, %	40.3	36.5	40.4	0.0332
Men, %	44.7	62.2	44.2	<0.0001
Body mass index, kg/m2	29.3(6.18)	29.4(6.13)	29.3(6.18)	0.8162
Current smoking, %	14.2	12.5	14.2	0.0205
History of cardiovascular disease,%	16.7	43.2	16.0	<0.0001
Family history of cardiovascular disease	57.2	58.7	57.2	0.4513
Diabetes, %	20.9	29.3	20.7	<0.0001
Hypertension, %	57.8	75.4	57.3	<0.0001
Atrial fibrillation, %	8.3	21.0	7.9	<0.0001
Left ventricular hypertrophy, %	9.6	20.1	9.3	<0.0001
Antihypertensive medication use, %	57.1	79.2	56.5	<0.0001
QT prolong drugs use, %	24.4	31.6	24.3	<0.0001
Warfarin use, %	3.1	11.0	2.9	<0.0001
Aspirin use,%	42.2	57.5	41.8	<0.0001
Systolic blood pressure, mmHg	127.3(16.55)	131.7(18.48)	127.2(16.47)	<0.0001
Heart rate, beats per min	66.7(11.39)	66.5(10.91)	66.71(11.40)	0.6989
QRS duration msec	87.4 (15.61)	123.8(25.63)	86.4(13.79)	<0.0001
Uncorrected QT interval, msec	395.6(32.80)	459.0(30.39)	393.9(31.07)	<0.0001
QTcFram interval, msec	407.2(23.30)	470.4(19.10)	405.4(20.84)	<0.0001

^*Continuous variables are reported as mean (SD)

^†Defined as values of 460 ms or longer in women and 450 ms or longer in men (12)

During up to 8.2 years of follow-up (median 5.1 years), 608 total strokes occurred, of which 491 were ischemic strokes. Approximately 5.4% of participants with prolonged QTc_Fram developed stroke compared to only 2.8% of those with normal QTc_Fram (p<0.0001). Figure 1 shows the event-free survival curves of participants with and without prolonged QTc_Fram.

Figure 1.

Kaplan Meir curves for incident stroke in REGARDS participants with and without prolonged QTc_Fram.

*The horizontal axis shows follow-up time in years, and the vertical axis the proportion of the population stroke-free for those without (red line) and with (blue line) prolonged QTc.

As shown in Table 2-A, for every 1 SD increase in QTc_Fram, there was 24% increase in the risk of incident stroke [(HR (95% CI): 1.24 (1.16, 1.33), p<0.0001)]. This association remained statistically significant after adjustment for demographics, traditional stroke risk factors, warfarin use, aspirin use, QRS duration and use of QT-prolonging drugs [HR (95% CI): 1.12(1.03,1.21), p=0.0053)]. Similar results were observed when QTc_Hod, QTc_Baz and QTc_Frid, separately, were used instead of the QTc_Fram in the proportional hazards models (Table 2-A).

Table 2.

Prolongation of QTc interval and risk of incident stroke in the REGARDS study

2-A: Hazards ratio (95% confidence interval) per 1-standard deviation increase in QTc
	Model 1*	p-value	Model 2†	p-value	Model 3‡§	p	Model 4II:	p
QTc-Framingham	1.24(1.16,1.33)	<0.0001	1.18(1.10,1.27)	<0.0001	1.11(1.04,1.20)	0.0036	1.12(1.03,1.21)	0.0053
QTc-Hodges	1.25(1.16,1.34)	<0.0001	1.18(1.10,1.26)	<0.0001	1.12(1.04,1.20)	0.0031	1.12(1.04,1.21)	0.0045
QTc-Bazett	1.26(1.18,1.36)	<0.0001	1.21(1.13,1.30)	<0.0001	1.13(1.05,1.21)	0.0009	1.14 (1.05,1.23)	0.0013
QTc- Fridericia	1.25(1.17,1.34)	<0.0001	1.19(1.11,1.27)	<0.0001	1.12(1.04,1.20)	0.0021	1.13(1.04,1.22)	0.0029

2-B: Hazards ratio (95% confidence interval) associated with prolonged QTc§(Ref=No)
	Model 1	p-value	Model 2*	p-value	Model 3†	p	Model 4‡§	p
QTc-Framingham	2.88(2.12,3.92)	<0.0001	2.08(1.52,2.84)	<0.0001	1.64(1.17,2.28)	0.0040	1.67(1.16,2.41)	0.0061
QTc-Hodges	2.45(1.79,3.34)	<0.0001	1.81(1.33,2.48)	0.0002	1.44(1.03,2.02)	0.0345	1.43(1.00,2.05)	0.0521
QTc-Bazett	2.02(1.55,2.63)	<0.0001	1.56(1.20,2.04)	0.0005	1.24(0.94,1.64)	0.1308	1.21(0.89,1.64)	0.2205
QTc- Fridericia	2.69(2.00,3.62)	<0.0001	1.95(1.45,2.63)	<0.0001	1.53(1.11,2.11)	0.0088	1.55 (1.09,2.21)	0.0148

^*Model 1, unadjusted

^†Model 2, adjusted for age, race, and sex

^‡Model 3, adjusted for model 1 plus traditional stroke risk factors (antihypertensive medication use, systolic blood pressure, current smoking, diabetes, left ventricular hypertrophy, atrial fibrillation, and prior cardiovascular disease)

^§Model 4, adjusted for model 3 plus warfarin use, aspirin use, QRS duration and use of QT-prolonging drugs

^IIProlonged QTc defined as 460 ms or longer in women and 450 ms or longer in men

The risk of stroke in study participants with prolonged QTc_Fram was almost three times the risk in those with normal QTc_Fram [HR (95% CI): 2.88 (2.12, 3.92), p<0.0001]. After adjustment for demographics, traditional stroke risk factors, warfarin use, aspirin use, QRS duration and use of QT-prolonging drugs, the risk of stroke remained significantly high [HR (95% CI): 1.67 (1.16,2.41), p=0.0061)]. Similar results were observed in the unadjusted and demographic adjusted models of the QTc_Hod, QTc_Baz and QTc_Frid. However, after further adjustment for stroke risk factors, warfarin use, aspirin use, QRS duration and use of QT-prolonging drugs, the hazard ratios associated with QTc_Baz were no longer statistically significant and were marginal with QTc_Hod (Table 2-B).

Figure 2 shows the multivariable-adjusted HRs for incident stroke associated with prolonged versus normal QTc_Fram in several subgroups of REGARDS participants stratified by age, race, sex, smoking status, hypertension, diabetes, LVH, AF and QRS duration. As shown, in a model adjusted for demographics, traditional stroke risk factors, warfarin use, QRS duration and use of QT-prolonging drugs, prolonged QTc_Fram (versus normal) was consistently associated with significantly increased stroke risk across all subgroups with non-significant probability values for interaction.

Figure 2.

Prolongation of QTcFram and risk of incident stroke within subgroups of REGARDS study participants

*Hazard ratios are adjusted for demographics (age, race, sex), traditional stroke risk factors (antihypertensive medication use, systolic blood pressure, smoking status, diabetes, left ventricular hypertrophy, atrial fibrillation, prior cardiovascular disease), warfarin use, QRS duration and use of QT-prolonging drugs

^†LVH, left ventricular hypertrophy; Atrial Fib, atrial fibrillation; CVD, prior cardiovascular disease

^‡ Black circles in the graph represent the HR and lines represent the 95% CI

^§ No significant interaction between the categories within each subgroups were observed.

Demographic adjusted “uncorrected” QT was marginally predictive of incident stroke [HR (95% CI): 1.08 (1.00, 1.16), p=0.0446 for every 1 SD increase, and 1.40 (1.01, 1.93), p=0.0423 for prolonged vs. normal]. However, these associations were no longer significant after further adjustment for traditional stroke risk factor [HR (95% CI): 1.06 (0.99, 1.15), p=0.1122 for every 1 SD increase, and 1.21 (0.86, 1.70), p=0.2709 for prolonged vs. normal] and further adjustment for warfarin use, aspirin use, QRS duration and use of QT-prolonging drugs [HR (95% CI): 1.06 (0.98, 1.15), p=0.1575 for every 1 SD increase, and 1.19 (0.84, 1.68), p=0.3367 for prolonged vs. normal]. Similarly, in models similar to those used for QTc, heart rate alone was not predictive of stroke after adjusting traditional stroke risk factors [HR (95% CI): 1.04 (0.96, 1.12), p=0.3309 for every 1 SD increase, and 0.97 (0.81, 1.15), p=0.7021 for abnormal vs. normal (60–90 beats per minute)] and further adjustment for warfarin use, aspirin use, QRS duration and use of QT-prolonging drugs [HR (95% CI): 1.04 (0.96, 1.12), p=0.3886 for every 1 SD increase, and 0.97 (0.81, 1.15), p=0.7182 for abnormal vs. normal (60–90 beats per minute)].

There were no meaningful differences in the observed patterns of associations assessing ischemic stroke only (data not shown). Also, there was no significant interaction between each of sex and race subgroups with QTc_Fram for prediction of stroke in the fully adjusted models (model 4).

DISCUSSION

In this analysis from a national, US general population, we showed that QTc prolongation is associated with a significantly increased risk of incident stroke independent of traditional stroke risk factors and consistently across different demographic and clinical subgroups. These findings add further concern regarding the consequences of QTc interval prolongation. In addition to increased cardiovascular morbidity and mortality (1–4), stroke seems to be another serious outcome. With the increasing number of QT-prolonging drugs (24), and in light of our results, examining the risk of stroke associated with QT-prolonging drugs may be warranted. To our knowledge, no previous studies examined the association between QT-prolonging drugs and stroke. Nevertheless, the Permanent Atrial Fibrillation Outcome Study Using Dronedarone on Top of Standard Therapy (PALLAS) Trial (25) gives an indirect clue of a possible link between QT-prolonging drugs and risk of stroke. In the PALLAS trial, stroke occurred in 23 patients in the dronedarone group and 10 in the placebo group (hazard ratio, 2.32; 95% CI, 1.11 to 4.88; P = 0.02). It was not clear how many of the stroke cases in PALLAS could be explained by the QT prolonging effect of dronedarone. It has been noticed, however, that after one month of treatment there was 8±40 msec in the dronedarone group and 2±38 msec in the placebo group (P<0.001), which accords with the known QT-prolonging properties of dronedarone. In contrary to the results of the PALLAS trial results, however, the ATHENA trial showed that dronedarone was associated with less stroke risk (26). It is not clear at this stage if the type of atrial fibrillation (being permanent in PALLAS and paroxysmal in ATHENA) plays a role in explaining such discrepancy. Another clue of a potential link between QT-prolonging drugs and risk of stroke comes from our study. In an additional analysis, we observed a 19% increase in the risk of stroke in participants on QT-prolonging drugs compared to those not on these drugs, after adjusting for age, sex and race [HR (95% CI): 1.19 (0.99, 1.42); p= 0.059]. However, this needs to be investigated further in a study designed specifically to address this question.

Two previous studies (9. 10), one from Brazil on diabetic population and another from Japan, showed significant relationship between prolongation of QTc and incident stroke, which accord with our results. However, both studies did not include whites (Caucasians) or blacks (African Americans), used only the debatable Bazett’s formula for calculating QTc, and at least one of them was lacking data on AF, smoking status and/or QT-prolonging drugs which makes it hard to reach a convening conclusion.

In our analysis, the association between QTc and stroke remained highly significant despite adjustment for traditional stroke risk factors. Noteworthy, we had a priori decision to focus on traditional stroke risk factors (i.e. components of the Framingham Stroke Risk Score) and major potential confounders. Nevertheless, in an additional analysis, we added other possibly related risk factors to model 4 including family history of CVD, chronic kidney disease and hypertension (as a categorical variable in addition to systolic blood pressure and antihypertensive medications which were already included in model 4). Further adjustment for these variables did not result in any significant change in the risk estimates (results not shown). This means that the association between QTc and CVD risk factors does not fully explain the prognostic significance of QTc as a stroke predictor. It is possible, however, that prolonged QTc interval is a marker of silent undetected atherosclerotic vascular disease (9). In one study QTc was significantly associated with carotid intima-media thickness (27) and in another study with activated factor XII levels (28). In both studies, the associations between QTc and these markers of atherosclerotic vascular disease remained significant after adjustment for other CVD risk factors. This reinforces the hypothesis that QTc prolongation may be a surrogate indicator of subclinical atherosclerosis, and subsequently can be predictive of future atherosclerotic vascular events such as stroke. Studies aimed to better clarify the mechanism by which QTc could be related to stroke risk are needed. Interestingly, the mechanistic relationship between stroke and LVH (which is one of the well established traditional stroke risk factors) is unclear too. All what we know is that LVH is associated with several CVD risk factors (which is the same situation with QTc). So, what has been said about LVH and risk of stroke (29) may also be applicable to QTc; that is “it is not yet entirely clear whether LVH (or QTc in our case) represents a marker, a limited adaptative process or a pathological process”

QT interval is heart rate dependent; the higher the heart rate the shorter the QT and vice versa. Therefore, using heart rate corrected QT (i.e. QTc) rather than the uncorrected QT is mandatory. This is further supported with our findings of lack of significant stroke risk with uncorrected QT while observing a strong association with QTc. Many formulas have been proposed to calculate QTc from the uncorrected QT interval and heart rate. The most widely used is the non-linear formula of Bazett (15). However, recent guidelines of ECG standardization and interpretation (12) recommend using linear regression formulas, avoiding non-linear formulas especially Bazett, and incorporating QRS duration in the models. Subsequently, we decided on using QTc_Fram and adjusting for QRS duration in the full model. Nevertheless, to confirm our results and to provide comparability with other studies, we also used three other QTc formulas including the traditional QTc_Baz. and the results were largely consistent. We also examined the association between heart rate and stroke in models similar to those we used for QTc, and we did not find significant associations. This precludes the possibility that the observed association between QTc and stroke was driven by heart rate.

As shown, the risk of stroke associated with 1 SD increase in QTc was consistent across the four QTc correction formulas we used in our study. However, when we examined the risk of stroke associated with prolonged QTc (versus normal), QTc_Baz was not as significantly predictive as other formulas in the multivariable models. This observation accords with the current general agreement that QTc_Baz can erroneously estimate QTc, which could affect its prognostic significance (12). In our study the prevalence of prolonged QTc (defined as QTc 460 ms or longer in women and 450 ms or longer in men) was approximately 5% by Bazett, and between 2.7 to 3% by the other three formulas. This means that at least 2% of our study participants were misclassified as having prolonged QTc by Bazett, which might have led to the null findings we observed. Lack of strong prognostic significance of QTc_Baz in comparison to the other QTc formulas supports the explicit recommendation of the current guidelines in avoiding the use of Bazett’s formula (12). Nevertheless, since the prognostic significance of QTc_Baz as a continuous variable (i.e. per 1 increase) was as good as other QTc formulas, using different (possibly higher cut-points) to define prolonged QTc_Baz may improve its prognostic significance. In other words, our findings suggest that the cut-points defining prolonged QT need to be formula-specific to avoid masking important relationships. Examining the impact of differences in the cut-points that define prolonged QTc on the prognostic significance is beyond the scope of this paper- which is the reason to decide on using the cut-points suggested by the current guidelines (12).

Our study has some limitations. Only whites and blacks were included in the REGARDS study, hence, our results may not be applicable to other race/ethnic groups. We could not differentiate between congenital and acquired (drug-induced) QTc prolongation. Nevertheless, given the age of REGARDS population and the higher prevalence of prolonged QTc in the users of QT-prolonging drugs, acquired QTc prolongation is a more likely etiology.

It remains possible that AF was under-detected in some cases. As antiarrhythmic drugs used in AF can often prolong the QT interval, one potential explanation for these findings may have been that more participants with undetected/unreported AF had drug-induced prolongation of QTc and subsequently an increased risk for stroke. However, given that AF in our study was ascertained by two methods [self-report of a previous physician diagnosis of AF and study scheduled ECG; both have similar stroke predictive value (22)], most of AF cases must have been captured. Also, we have adjusted for QTc prolonging drugs in our analysis, which preclude the possibility of the confounding effect of antiarrhythmic drugs with QTc-prolonging properties.

Data on use of medications (from which we identified QT-prolonging drugs) is based on review of pill bottles brought by participants during the study baseline. Reporting medications by the participants may have been incomplete or may have changed prior to the incident event leading to misclassification. However, such inaccuracies or misclassification should be randomly distributed across groups and should not affect the overall conclusions.

Despite these limitations, this is the first report examining the relationship between QTc prolongation and risk of stroke in one of the largest biracial population-based longitudinal cohort studies in the US; REGARDS. Other strengths include central ECG reading, using multiple QTc correction formulas to confirm the results, and the substantial accumulating number of physician-adjudicated stroke events. In conclusion, prolongation of QTc is associated with a significantly increased risk of incident stroke independently from traditional stroke risk factors. These findings suggest that important predictive information may be derived from such a simple ECG marker. With the increasing number of QT-prolonging drugs, and in light of our results, examining the risk of stroke associated with QT-prolonging drugs may be warranted.

List of abbreviations

REGARDS

the REasons for Geographic and Racial Differences in Stroke study

QTc_Fram

Framingham heart rate corrected QT

QTc_Hod

Hodge’s heart rate corrected QT

QTc_Frid

Fridericia’s heart rate corrected QT

QTc_Baz

Bazett’s heart rate corrected QT

LVH

Left ventricular hypertrophy

AF

Atrial fibrillation