Optimal QT Correction Formula for Older Chinese: Guangzhou Biobank Cohort Study

Electrocardiogram (ECG) is a noninvasive test widely used in health checks globally. The QT interval is an important parameter on ECG. QT prolongation reflects abnormalities in cardiac repolarization and is a risk marker for ventricular arrhythmias and sudden cardiac death [1]. Some prospective studies showed that prolongation of the QT interval was associated with higher risks of all-cause mortality, cardiovascular disease (CVD) mortality, and incident CVD in the general population [2‒5].

As the QT interval decreases while the heart rate increases, the QT interval is usually normalized with respect to the heart rate, and it is denoted as corrected QT (QTc) [6]. Many formulas have been proposed to calculate QTc. The most widely used formula is the Bazett formula, although it tends to provide overcorrection value when used at high heart rates and under-correction value at low heart rates [7]. Several other correction formulas such as the Fredericia formula [8], Framingham formula [9], and Hodges formula [10], have been used in different settings. Although previous studies have compared these formulas in patients with atrial fibrillation [11], sinus tachycardia [12], and other diseases [13], QT interval can be prolonged by medications, electrolyte abnormalities and ischemia, and their findings might not be directly applicable to the general population. In addition, the performances of the formulas are different as heart rate varies [14], and thus the cutoff for prolonged QTc should be population- and formula-specific [15]. Notably, the current cutoffs for prolonged QTc (450 millisecond [ms] for men and 460 ms for women) were proposed by the American Heart Association, American College of Cardiology, and Heart Rhythm Society based on percentiles of QTc corrected by linear regression formulas (such as Framingham formula) [15]. We found no reports on which correction formula and cutoffs for prolonged QTc are applicable to Chinese. It is important to identify the appropriate correction formula and cutoffs for prolonged QTc specifically for the Chinese population, especially given the comprehensive coverage of ECGs under China National Essential Public Health Services Program for older adults aged 65 years or above [16]. Using data from the Guangzhou Biobank Cohort Study (GBCS), we aimed to (1) identify the optimal QT correction formula for generating the QTc, and the optimal cutoffs for prolonged QTc; (2) examine the associations of prolonged QTc defined by new cutoffs with all-cause mortality, CVD mortality, and incident CVD in older Chinese.

In GBCS at baseline, 30,430 participants aged 50+ years were recruited from November 2003 to January 2008. Details of GBCS have been reported previously [17]. Participants were recruited from “The Guangzhou Health and Happiness Association for the Respectable Elders” (GHHARE), a community social and welfare organization. GHHARE included about 7% of Guangzhou residents in this age group, with branches in all districts of Guangzhou, the capital city of Guangdong Province in southern China. The Guangzhou Medical Ethics Committee of the Chinese Medical Association approved the study (No. 2020044). All participants gave written informed consent before participation.

Of the 30,430 participants at baseline, 5,819 participants were excluded, including 2,609 participants with incomplete information on QT interval and heart rate, 917 participants with atrial fibrillation, bundle branch blocks and implausible heart rate (>200 beats/minute), and 2,293 with a self-reported history of CVD (coronary heart disease [CHD], stroke, angina, myocardial infarction [MI], peripheral vascular disease, heart failure [HF], and congenital heart disease), giving 24,611 participants in the present analyses.

ECG was performed in the supine position after resting for 5 min [18]. A standard ECG was performed using a 3-channel, 12-lead electrocardiograph (Marquette MAC-500) during phase 1 and at the initial part of phase 2. A synchronous 12-lead ECG (Marquette Cam-14 acquisition module) was used for the remainder of phase 2 and in phase 3 [19]. QT intervals in phase 1 and the beginning of phase 2 were measured manually by two qualified physicians, who were blinded to other participant information, with any discrepancies resolved through consensus [18‒20]. For the remainder of phase 2 and in phase 3, QT intervals were measured automatically by the ECG device [19, 20]. Four most common formulas for QT correction were used, as follows: (1) Bazett [7] $QTb=QT/RR$⁠; (2) Fridericia [8] $QTfri=QT/RR3$⁠; (3) Framingham [9] $QTfra=QT+154×1−RR$⁠; (4) Hodges [10] $QTh=QT+105×1/RR−1$⁠.

Outcomes included all-cause mortality, CVD mortality and the first development of nonfatal or fatal CVD. Causes of death and CVD were coded according to the 10th revisions of the International Classification of Diseases (ICD-10) by trained clinical coding officers in each hospital. CVD was defined as any hospital admission or death from CHD (I20-I25), stroke (I60-I69), peripheral artery disease (I73), and HF (I50). When a participant had multiple CVD events, the first event was designated as the incident event. Information on underlying causes of deaths up to July 2022 was mostly obtained via record linkage with the Death Registry Department of the Guangzhou Centre for Disease Control and Prevention (GZCDC). Incidence information on CVD up to December 2020 was collected from the hospitalization data of the Guangzhou Social Insurance Bureau and GZCDC. The methods have been reported elsewhere [21, 22].

Potential confounders included sex, age (continuous), education (primary/middle school/college), occupation (manual/non-manual/other), annual household income (<30,000 RMB per year/≥30,000 RMB per year/unknown), smoking status (never/former/current), alcohol use (never/former/current), physical activity (inactive/moderate/active), hypertension, diabetes, left ventricular hypertrophy, body mass index, and triglyceride. These methods for measurement are described in online supplementary material (for all online suppl. material, see https://doi.org/10.1159/000542238).

Linear regression was used to regress QTc on heart rate to calculate the slope and the intercept with the 95% confidence interval (CI). The QT correction formula with the slope of the regression line and R² closest to zero would be considered as the best because it would be least likely to be affected by heart rate. We used the 95th percentile to determine the cutoffs for QTc to define prolonged QTc. The sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were calculated for predicting 10-year risk of all-cause mortality, CVD mortality, and incident CVD. Chi-square tests or analysis of variance were used to compare participants’ baseline characteristics by the presence of prolonged QTc. Cox regression was used to estimate the associations of prolonged QTc with all-cause mortality, CVD mortality, incident CVD and specific CVD. The proportional hazards assumption was checked by visual inspection of plots of log (−log S) against time using “stphplot” command in STATA, where S was the estimated survival function. Censor date was defined as the date of diagnosis of CVD, death, or end of the follow-up in this study (July 31st, 2022), whichever came first. We conducted likelihood ratio tests to compare fitness of models with and without including interaction terms between prolonged QTc and sex. The continuous net reclassification index (NRI) was calculated to assess the reclassification for 10-year risk of all-cause mortality, CVD mortality, and incident CVD after adding prolonged QTc to the baseline model. The baseline model included variables within Framingham Risk Score (sex, age, diabetes, smoking status [never, ever], systolic blood pressure, total cholesterol, and high-density lipoprotein cholesterol). The NRI was calculated using R package survIDINRI.

Although serum kalium data were available for only 1,733 participants, we further adjusted for hypokalemia (<3.5 mmol/L) in sensitivity analyses, as electrolyte disturbances such as hypokalemia can significantly affect the QT interval. Additionally, as a sensitivity analysis, we also used the 98th percentile to define prolonged QTc and conducted the same analysis as with the 95th percentile. Statistical analyses were done using STATA/MP 17.0 and R 4.2.1. All tests were two-sided with p < 0.05 as statistically significant.

Online supplementary Figure 1 shows that all four QTc distributions were narrower than the uncorrected QT distribution. The QTc corrected by Bazett formula was relatively wider and was an outlier compared to all other formulas results. Table 1 shows that the mean heart rate, QTb, QTfri, QTfra, and QTh for all the participants were 72.72 (standard deviation 10.90) beats/minute, 422.97 (25.39) ms, 410.27 (22.66) ms, 410.85 (21.38) ms, and 408.94 (21.50) ms, respectively. Women had approximately 10 ms longer values than men for the average QTc (Table 1).

Table 2 shows that the regression coefficients (β (95% CI)) between heart rate and QTc, as well as the R², were the lowest for the model using the Hodges formula (QTh): for men 0.041 (−0.002 to 0.084), R² 0.001; for women −0.06 (−0.08 to −0.03), R² 0.001; for all participants −0.001 (−0.026 to 0.023), R² < 0.001. In contrast, QTc corrected by Bazett formula (QTb) showed the strongest association with heart rate (β [95% CI]: men 0.99 [0.95 to 1.04]; women 0.82 [0.79 to 0.85]; all participants 0.90 [0.87 to 0.93]) and the highest R² values (men 0.22; women 0.12; all participants 0.15).

Table 3 shows that in men, the cutoff based on the 95th percentile of QTc was 455.67 ms for QTb, 437.89 ms for QTfri, 437.04 ms for QTfra, and 435.75 ms for QTh, and in women, the cutoff was 464.81 ms, 448.14 ms, 447.30 ms, and 445.25 ms, respectively. In men, the cutoff based on the 98th percentile of QTc was 465.53 ms for QTb, 446.86 ms for QTfri, 445.67 ms for QTfra, and 444.50 ms for QTh, and in women, the cutoff was 475.63 ms, 458.39 ms, 456.09 ms, and 454.54 ms, respectively.

Women had approximately 10 ms longer values than men for the cutoffs. The cutoffs for QTb in both men and women had approximately 20 ms longer values than those for QTfri, QTfra, and QTh. In addition, the 98th percentile values for QTfri, QTfra, and QTh in both men and women were close to the sex-specific clinical standards (450 ms for men and 460 ms for women), while those for QTb were significantly higher than clinical standards.

Online supplementary Table 1 shows that compared to participants with normal QTc at baseline, those with prolonged QTc (≥95th percentile) were older, had higher proportion of lower education, hypertension and left ventricular hypertrophy, and higher levels of systolic blood pressure, diastolic blood pressure, body mass index, and triglycerides. In addition, compared to participants with normal QTb, those with prolonged QTb had higher proportion of diabetes, higher levels of fasting plasma glucose, low-density lipoprotein cholesterol, and lower level of high-density lipoprotein cholesterol. Participants with prolonged QTfri, QTfra, and QTh had lower level of total cholesterol than those with normal QTc.

During an average follow-up of 15.3 (standard deviation 3.4) years, 5,261 deaths and 5,539 CVD were recorded. Table 4 shows that after adjusting for 13 potential confounders, prolonged QTc (≥95th percentile) remained an independent risk factor for all-cause mortality and CVD mortality in all participants. The prolonged QTc corrected by the Bazett formula appeared to have the strongest association with higher risks of all-cause mortality (hazard ratio [HR] 1.40 [95% CI 1.25–1.56]) and CVD mortality (HR 1.52 [1.25–1.84]) in all participants. The prolonged QTc corrected by the Fridericia, Framingham, and Hodges formulas had a similar association with higher risks of all-cause mortality (QTfri: HR 1.25 [1.11–1.41], QTfra: HR 1.26 [1.12–1.42], QTh: HR 1.22 [1.08–1.37]) and CVD mortality (QTfri: HR 1.41 [1.15–1.72], QTfra: HR 1.42 [1.16–1.74], QTh: HR 1.36 [1.12–1.66]). The prolonged QTb was not associated with a higher risk of incident CVD (HR 1.10 [0.97–1.24]) in all participants, while prolonged QTfri (HR 1.15 [1.02–1.30]), QTfra (HR 1.19 [1.05–1.34]) and QTh (HR 1.14 [1.01–1.29]) remained associated with a higher risk of incident CVD.

Table 5 shows that, in all participants, prolonged QTb was associated only with a higher risk of ischemic stroke (HR 1.37 [1.12–1.67]), while prolonged QTfri, QTfra, and QTh were associated with higher risks of CHD, MI, ischemic event (CHD or ischemic stroke), and ischemic stroke. Compared to normal QTc, prolonged QTc (QTfri, QTfra, and QTh) were associated with approximately 26%, 50%, 20%, and 40% increased risks of CHD, MI, ischemic event, and ischemic stroke, respectively.

Further, for all participants, we found no evidence that the associations of prolonged QTc with all-cause mortality, incident CVD and specific CVD varied by sex (Tables 4, 5). However, significant interactions between prolonged QTc (QTfri, QTfra, and QTh) and sex on CVD mortality were found (p for interaction from 0.003 to 0.03) (Table 4). The associations of QTfri (HR 1.82 [1.43–2.31] vs. 0.92 [0.63–1.34]), QTfra (HR 1.74 [1.36–2.23] vs. 1.07 [0.75–1.52]), and QTh (HR 1.68 [1.32–2.13] vs. 0.98 [0.68–1.40]) with CVD mortality were found in women, but not in men.

Table 4 shows that the cutoffs for prolonged QTb, QTfri, QTfra, and QTh based on the 95th percentile showed similar sensitivity, specificity, PPV, and NPV in predicting risk of all-cause mortality, CVD mortality, and incident CVD. In addition, adding prolonged QTb to a model containing traditional risk factors did not improve 10-year all-cause mortality (NRI 0.065 [−0.011 to 0.097]), CVD mortality (NRI 0.062 [−0.138 to 0.150]), and incident CVD (NRI 0.069 [−0.035 to 0.097]) risk reclassification in all participants. The addition of prolonged QTfri, QTfra, and QTh improved risk reclassification by approximately 5% for all-cause mortality, approximately 10% for CVD mortality, and approximately 6% for incident CVD (Table 4).

Online supplementary Table 2 shows that, in a subset of 1,733 participants, after further adjusting for hypokalemia, the associations of prolonged QTc with all-cause mortality, CVD mortality, and ischemic stroke were no longer statistically significant, possibly due to the small sample size. However, associations with incident CVD, CHD, MI, and stroke remained, especially for QTfri, QTfra, and QTh. Significant interactions between sex and these formulas for incident CVD, CHD, MI, and ischemic events were also observed (p for interaction from <0.001 to 0.04). Online supplementary Table 3 shows significant associations of QTc with CVD, CHD, MI, and ischemic event in women, but not in men.

We also used the 98th percentile to define prolonged QTc. Online supplementary Table 4 shows that prolonged QTc was associated with higher risks of all-cause mortality and CVD mortality but not with a higher risk of incident CVD. Online supplementary Table 5 shows that prolonged QTc was associated with higher risks of CHD, MI, ischemic event, and ischemic stroke. These associations with mortality from all-cause and CVD, incident CVD, and specific CVD did not vary by sex (online suppl. Tables 4, 5, p for sex interaction ≥0.05).

Online supplementary Table 4 shows that the cutoffs for prolonged QTb, QTfri, QTfra, and QTh based on the 98th percentile had similar sensitivity, specificity, PPV and NPV in predicting risk of all-cause mortality, CVD mortality, and incident CVD. Although prolonged QTc defined by the 98th percentile was associated with higher risks of all-cause mortality and CVD mortality, adding them to the baseline model containing traditional risk factors did not improve the reclassification for all-cause mortality, CVD mortality, and incident CVD (p for NRI ≥0.05).

To the best of our knowledge, our study for the first time evaluated various QT interval correction formulas in older Chinese. We found the Fridericia, Framingham, and Hodges formulas had similar performance in mathematical heart rate correction, with the Hodges formula performing the best. Prolonged QTc corrected by these three formulas showed similar associations and predictive performance with all-cause mortality, CVD mortality, and incident CVD, especially in women. They were mainly associated with higher risks of CHD, MI, and ischemic stroke. However, the Bazett formula performed the worst in both heart rate correction and risk prediction.

Our results were consistent with previous studies comparing different formulas of QTc (QTb, QTh, QTfra, and QTfri) [12‒14, 23], showing that the Bazett formula had the strongest association with heart rate and it overestimated the number of participants with a prolonged QTc. For example, the Penn Atrial Fibrillation Free Study on 6,723 patients without a history of HF and with baseline sinus rate >100 beat/minute showed that if using the QTc corrected by the Bazett formula, 39% would be classified as prolonged QTc defined by the current clinical standards (450 ms for men and 460 ms for women), which was much higher than those diagnosed by Fridericia, Framingham, and Hodges formulas (<9% being classified as prolonged QTc) [12]. Similarly, in our study, the 98th percentile cutoffs for QTfri, QTfra, and QTh in both men and women were close to the sex-specific clinical standards (450 ms for men and 460 ms for women), while those for QTb were significantly higher than clinical standards (Table 3). Thus, using the current clinical standards leads to an overestimation of patients with prolonged QTc when using Bazett formula.

In our study, the prolonged QTc (≥ the 95th percentile) corrected by Bazett formula was not associated with a higher risk of incident CVD, and did not improve the predictive performance of the baseline model with variables within Framingham Risk Score for all-cause mortality, CVD mortality, and incident CVD. The Penn Atrial Fibrillation Free Study also showed prolonged QTc defined by Bazett formula was not associated with higher risks of mortality and the association with incident CVD was the weakest among the four formulas [12]. Additionally, all prolonged QTc defined by the 98th percentile was not associated with a higher risk of incident CVD, nor did it improve risk prediction for mortality and CVD. Thus, our findings do not recommend using the Bazett formula and the current clinical standards for defining prolonged QTc in older Chinese.

In contrast, the prolonged QTfri, QTfra, and QTh defined by the 95th percentile was not only associated with higher risks of death and incident CVD, but also improved risk reclassification by approximately 5%, 10%, and 6% for all-cause mortality, CVD mortality, and incident CVD, respectively (Table 4). The prolonged QTfri, QTfra, and QTh were particularly associated with higher risks of ischemic events including CHD, MI and ischemic stroke. However, the Tromsø Study, a Shanghai study, and a Liaoning study showed prolonged QTc was not associated with a higher risk of CHD [24] or MI [25, 26], likely due to participants were young with an average age of 36.3 years old [25], and low incidence and insufficient statistical power [24, 26]. Additionally, the formula used in the Tromsø Study and Liaoning study was the Bazett formula, and the prolonged QTc was defined by the clinical standards (450 ms for men and 460 ms for women) [24, 25]. Our study also showed that prolonged QTc corrected by the Bazett formula was not associated with higher risks of CHD and MI (Table 5). Studies on the associations of QTc with stroke were reported [5, 24, 26, 27]. Notably, these studies did not differentiate between stroke types. In our study, prolonged QTc corrected by the four formulas and defined by the 95th percentile was associated with a higher risk of ischemic stroke, with being HR approximately 1.40 (Table 5), while the association of prolonged QTc with hemorrhagic stroke was not found. Previous studies have shown that prolongation of the QT interval is associated with the occurrence of early afterdepolarization, in which abnormal depolarization occurs during phases two or three of the action potential before repolarization is completed [28]. These premature or triggered action potentials can generate cardiac arrhythmias such as torsade de pointes, which may progress to ventricular fibrillation, myocardial ischemia, and infarction [29]. This may in part explain the higher ischemic event risk associated with prolonged QTc.

Another finding from the present analysis is the lack of associations of prolonged QTc with CVD mortality in men (p for sex interaction from 0.003 to 0.03) (Table 4). Although no interaction between prolonged QTc (QTfri, QTfra, and QTh) and sex on incident CVD was found, the associations in men were not statistically significant. Testosterone has been shown to shorten the action potential duration, by enhancing the expression of K⁺ channels and downregulating I_CaL increasing the repolarization reserve of the cardiac muscle [30]. In addition, testosterone replacement therapy can improve many CVD risk factors and slow the progression of atherosclerosis [31], which might weaken the association of prolonged QTc with CVD in men.

The results of previous studies regarding heart rate correction and the independent predictive value of QTc formulas were inconsistent, possibly due to differences in the cutoffs used, heterogeneous study samples, and/or various duration of follow-up [32]. In our study, although Hodges formula showed better heart rate correction than Framingham and Fridericia formulas in our study, the differences were small, and the prolonged QTc corrected by these three formulas had similar associations and predictive performance with all-cause mortality, CVD mortality, and incident CVD. It has been shown that in resting heart rates of 60–90 beats/minute, different formulas tend to provide similar results for detecting QT prolongation [14, 32]. As 90.22% of participants in the present study had heart rates in this range, the impact of the correction formula may have been relatively minor.

Our study had some limitations. First, the participants were recruited from the GHHARE, and like all community-based cohorts, they were relatively healthy. Selection and survival biases could not be completely ruled out and the HRs could be underestimated. Second, the QT interval was measured at once baseline, which might be subject to errors due to within-individual variability. Using repeated measurements of QT interval to obtain individualized QTc may reduce such errors but is not practical in large epidemiologic studies. Third, QT intervals were measured manually by two qualified physicians during the early participant recruitment phases and automatically by the ECG device during the later phases. While manual measurement may introduce variability, we believe any potential measurement errors were likely to be random, which could be mitigated by our large sample size. Additionally, the QT intervals were independently measured by two physicians who reached a consensus, minimizing potential discrepancies. Fourth, another limitation of this study is the lack of data on the use of antiarrhythmic drugs, antibiotics, antipsychotic drugs, which can impact QT interval duration. Although we adjusted for hypokalemia in sensitivity analyses, further studies are warranted to validate our findings in light of these unmeasured factors. Finally, our participants were urban Chinese and might not be representative of the broader population, whether our findings can be generalized to other ethnic groups and other settings is uncertain.

We found that Hodges formula outperformed the other formulas for correcting heart rate. Prolonged QTc corrected by Fridericia, Framingham, and Hodges formulas and defined by the 95th percentile can be used for death and cardiovascular risk prediction.

This study protocol was reviewed and approved by the Guangzhou Medical Ethics Committee of the Chinese Medical Association, approval No. 2020044. All participants gave written informed consent before participation.

The authors have no conflicts of interest to declare.

This work was supported by the National Natural Science Foundation of China (Grant No. 82373661) and China Scholarship Council. The Guangzhou Biobank Cohort Study was funded by the University of Hong Kong Foundation for Educational Development and Research (Grant No. SN/f/HKUF-DC; C20400.28505200); the Health Medical Research Fund (Grant No. HMRF/13143241) in Hong Kong; the Guangzhou Public Health Bureau (Grant No. 201102A211004011); the Natural Science Foundation of Guangdong (Grant No. 2018A030313140); and the University of Birmingham, UK. The funder had no role in the design, data collection, data analysis, and reporting of this study.

W.B.T., W.S.Z., C.Q.J., X.Y.L., F.Z., Y.L.J., T.Z., T.H.L., K.K.C., and L.X. have substantial contributions to conception and design, acquisition of funding, data and interpretation of data; W.B.T.: formal analysis; W.B.T., L.X., T.H.L., and K.K.C.: writing – original draft; W.S.Z., C.Q.J., X.Y.L., F.Z., Y.L.J., and T.Z.: resources. All authors contributed to final approval of the paper.

The datasets analyzed during the current study are not publicly available due to the protection of the privacy of participants but are available from the corresponding author on reasonable request.