Abstract
Background
The prognostic value of T‐wave morphology parameters in coronary artery disease in the current treatment era is not well established.
Methods
The Innovation to reduce Cardiovascular Complications of Diabetes at the Intersection (ARTEMIS) study included 1,946 patients with angiographically verified coronary artery disease (CAD). The study patients underwent thorough examinations including 12‐lead digital electrocardiogram (ECG) at baseline.
Results
During a follow‐up period of 73 ± 22 months, a total of 201 (10.3%) patients died. Of the study patients, 95 (4.9%) experienced cardiac death (CD) consisting of 44 (2.3%) sudden cardiac deaths (SCD) and 51 (2.6%) nonsudden cardiac deaths (NSCD), and 106 (5.4%) patients experienced noncardiac death (NCD). T‐wave morphology dispersion (TMD), T‐wave area dispersion (TWAD), and total cosine R‐to‐T (TCRT) had a significant association with CD even after adjustment with relevant clinical risk markers in the Cox regression analysis (multivariate HRs: 1.015, 95% CI 1.007–1.023, p = .0003; 0.474, 95% CI 0.305–0.737, p = .0009; 0.598, 95% CI 0.412–0.866, p = .006, respectively). When including these parameters to the clinical risk model for CD, the C‐index increased from 0.810 to 0.823 improving the discrimination significantly (integrated discrimination index [IDI] = 0.0118, 95% CI 0.0028–0.0208, p = .01). These parameters were more closely associated with NSCD (multivariate p‐values from .016 to .001) than with SCD (univariate/multivariate p‐values for TMD .015/.197 and for TCRT .012/.43).
Conclusion
T‐wave morphology parameters describing repolarization heterogeneity improve the predictive power of the clinical risk model for CD in patients with CAD in the current treatment era.
Keywords: electrocardiogram, repolarization, T‐wave morphology
1. INTRODUCTION
Coronary heart disease is the single most important cause of mortality accounting for approximately 30% of all deaths, out of which sudden cardiac death (SCD) accounts for over 50% (Adabag, Luepker, Roger, & Gersh, 2010; Deo & Albert, 2012; Engdahl, Holmberg, Karlson, Luepker, & Herlitz, 2002; Zipes & Wellens, 1998). The risk of CD and SCD can only very roughly be evaluated by general clinical characteristics and left ventricular function (Goldberger et al., 2011; Zipes & Wellens, 1998). Furthermore, other methods used for prognostic purposes also lack good accuracy (Verrier & Huikuri, 2017). During the past decades, three‐dimensional vectorcardiography has gained popularity in studying spatiotemporal changes in electrophysiology of myocardium (Verrier & Huikuri, 2017). T‐wave morphology parameters have been shown to yield prognostic information in various populations (Kardys et al., 2003; Kors et al., 1998; Perkiömäki et al., 2006; Porthan et al., 2013; Zabel et al., 2000; Zabel et al., 2002) . Studies in postinfarction patients with well‐preserved left ventricular function have suggested that some of these parameters are more closely associated with risk of arrhythmic events and cardiac mortality than total mortality (Perkiömäki et al., 2006; Zabel et al., 2000). Substantial advancements have been made in the treatment of patients with coronary artery disease (CAD) and acute coronary syndromes in the past decade which have improved the prognosis of these patients (Piironen et al., 2017). However, the prognostic value of T‐wave morphology parameters in patients with CAD in the current treatment era is not well established. Therefore, the present study was aimed to test the hypothesis that T‐wave morphology parameters evaluating repolarization heterogeneity predict CD in patients with CAD in the modern treatment era. To test the hypothesis, we assessed the prognostic value of several T‐wave morphology parameters in 1,946 patients with angiographically verified CAD during over 6 years of follow‐up.
2. METHODS
2.1. Study population
The Innovation to reduce Cardiovascular Complications of Diabetes at the Intersection (ARTEMIS) study included 1,946 patients with angiographically verified CAD defined as at least 50% stenosis in one or more major coronary vessels between August 2007 and November 2011. The risk assessments were done from 3 to 6 months after the coronary angiography. The study was performed at the Division of Cardiology of the University Hospital of Oulu (NCT01426685). Of the patients 833 had diabetes. Diabetes mellitus type 2 was defined according to guidelines of WHO (2006). The Artemis study was conducted according to the recommendations of Declaration of Helsinki, and protocol was approved by the ethical committee of Northern Ostrobothnia Hospital District. An informed consent was required from all subjects. Exclusion criteria in the study were age under 18 or over 80 years, New York Heart Association (NYHA) class IV or Canadian Cardiovascular Society class IV, planned or existing implantable‐cardioverter defibrillator, life expectancy <1 year, pregnancy, end‐stage renal dysfunction requiring dialysis, and being otherwise unfit for the study due to physical or psychological reasons, or due to doubted compliance. Patients were under regular control due to CAD, and questionnaires and telephone calls were performed during the follow‐up to receive prevailing data. Assessing risk markers for cardiac events and the influence of treatments on reducing cardiovascular complications in patients who have angiographically documented CAD with or without diabetes is among the main aims of the ARTEMIS study.
2.2. Endpoints of the present study
Mode of death was divided into two main groups; CD or noncardiac death (NCD). Patients were considered experienced CD if no apparent reason for NCD was found. Further division was done within CD into SCD which was defined as witnessed death within 1 hr from the onset of symptoms, or death occurring within 24 hr of last witnessed moment being alive, if no proof of symptoms’ onset was available. Nonsudden cardiac death (NSCD) included the patients classified having experienced CD, but not meeting the criteria to be included in SCD group.
2.3. Electrocardiography and T‐wave morphology parameters
Digital 12‐lead ECGs were obtained from 3 to 6 months after the coronary angiography. Collected ECG data were processed and T‐wave morphology parameters were calculated using custom‐made ECG‐analysis software. All ECG parameters were calculated automatically using methodology described earlier in detail (Acar, Yi, Hnatkova, & Malik, 1999). T‐wave loop dispersion (TWLD) is established by setting the loop inside a rectangle in the preferential plane. The rectangle is divided into 10 × 10 (100) subdivisions of smaller rectangles. TWLD is defined as the number of the smaller rectangles that are touched with the T‐wave loop outline (Perkiömäki et al., 2006). The first dimension of the principal component analysis of the T‐wave loop (T_PCA1) describes the length of the loop and the second dimension the height of the loop in the preferential plane (T_PCA2). The third dimension (T_PCA3) describes three‐dimensional deviation of the loop from the two‐dimensional preferential plane. T‐wave morphology dispersion (TMD) expresses mainly the variation of T‐wave between individual leads (leads I–II and V2–V6) (Porthan et al., 2013). The analysis of variation of T‐wave morphology between individual leads is based on calculating the average angle between all reconstruction vector pairs in T‐wave loop, higher values being an indicator of abnormal variation in repolarization (Porthan et al., 2013). TCRT stands for total cosine R‐to‐T, the angle formed between QRS‐complex loop and T‐wave loop vectors reflecting the spatial angle between depolarization and repolarization (Kenttä, Viik, & Huikuri, 2012; Porthan et al., 2013). Smaller values indicate larger angle between depolarization and repolarization fronts and are considered abnormal (Smetana, Batchvarov, Hnatkova, Camm, & Malik, 2002). The corresponding angle is also described using the mean spatial QRS‐T angle (QRSTmean), in which the loops’ X‐, Y‐ and Z‐ projection areas are applied for angle calculations. T‐wave area dispersion (TWAD) is obtained from leads V4–V6, where the average T‐wave area is divided by the maximum absolute T‐wave area within leads V4 to V6, which results in smaller values for highly heterogeneous T‐wave area (Kenttä et al., 2015). TLL3D stands for the T‐wave loop length measured in three‐dimensional space. QT interval corrected for heart rate by the Bazett's formula (QTc) and QT dispersion (QTd) calculated by subtracting the shortest QT interval from the longest among the 12 leads of ECG were also measured from the ECG.
2.4. Echocardiography
Echocardiographic examinations were performed using the General Vivid 7 ultrasound instrument (General Electric Healthcare, Little Chalfont, UK) with both examination and analysis conducted according to the general guidelines of American Society of Echocardiography by three cardiologists (Lang et al., 2015). Left ventricular ejection fraction (LVEF) was determined using 2D method. Left ventricle mass index (LVMI) was obtained by calculating LVM (left ventricle mass) and dividing it by body surface area (BSA). LVM was established using standardized formula for LVM; 0.8 × 1.04 × [(IVS + LVEDD + PWT)3−LVEDD3] + 0.6, where IVS stands for interventricular septum, LVEDD for left ventricular end‐diastolic diameter, and PWT for inferolateral/posterior wall thickness with all measurements derived at end‐diastole (Lang et al., 2015).
2.5. Statistical analysis
The standard t test was used to assess the statistical significance of the differences between continuous variables and the chi‐square test between categorical variables. The clinical variables that differed significantly in univariate comparisons were tested in the multivariate Cox regression analysis using stepwise forward analysis. The T‐wave morphology parameters that differed significantly in univariate comparisons were tested in this model one at a time as continuous variables. Case‐specific follow‐up times were applied for each patient individually in time‐specific analyses. The C‐index and integrated discrimination index (IDI) were calculated to assess the improvement of discrimination accuracy of the risk model for CD when T‐morphology parameters were added in the model. Reclassification accuracy was evaluated by the net reclassification index (NRI). The parameters were added in the model as continuous variables. The Kaplan–Meier curves were used to show the cumulative proportional probabilities of different modes of death. The log‐rank test was used to assess the statistical significance of the separation of the curves. The cut points were optimized from receiver operating characteristics (ROC) curves at the sensitivity level from 25% to 50%. Analyses were conducted using IBM SPSS version 24 (IBM SPSS, Armonk, NY, USA). A p‐value <.05 was considered significant.
3. RESULTS
During a follow‐up period of 73 ± 22 months, a total of 201 (10.3%) patients died. Of the study patients, 95 (4.9%) experienced CD consisting of 44 (2.3%) SCD and 51 (2.6%) NSCD cases, and 106 (5.4%) patients experienced NCD.
3.1. Association of baseline factors with cardiac, sudden cardiac, nonsudden cardiac, and noncardiac death
The patients who experienced CD during the follow‐up were older, more often males, had more commonly diabetes, consumed more alcohol, had worse NYHA class, had more commonly claudication, were less commonly of cholesterol‐lowering medication, more commonly on diuretic, and long‐acting nitrate medication, had higher LVMI, lower LVEF, larger LVEDD, and larger left ventricular end‐systolic diameter (LVESD) compared with patients who remained alive (Table 1). The patients who died because of cardiac causes had higher heart rate, longer QTc, lower values of T_PCA1, TCRT and TWAD, and higher values of TMD and QRSTmean compared with patients who remained alive (Table 2). The associations of SCD, NSCD, and NCD with clinical and T‐morphology parameters are also shown in Table 1 and 2. After adjustments in a stepwise forward multivariate Cox regression analysis, higher age, diabetes, higher consumption of alcohol, worse NYHA class, more common usage of long‐acting nitrates, lower LVEF, and larger LVESD still retained a significant association with the risk of CD (Table 3); higher age, diabetes, higher consumption of alcohol, more common usage of long‐acting nitrates, and lower LVEF with the risk of SCD (Table 3); higher age, diabetes, less common usage of cholesterol‐lowering medication, more common usage of long‐acting nitrates, higher LVMI, and lower LVEF with the risk of NSCD (Table 3); and higher age, male gender, former smoking, claudication, more common usage of diuretic medication, and negative family history of SCD with the risk of NCD (Table 3). When the T‐wave morphology parameters which had significant association with the corresponding mode of death in univariate comparisons were tested one at a time in the clinical multivariate Cox hazards model, higher values of TMD, and lower values of TCRT and TWAD retained their significant association with CD (Table 3); and higher values of TMD, and lower values of TCRT and TWAD with NSCD (Table 3). Longer QTc retained significant association with NCD, however, none of the parameters retained significant association with SCD (Table 3). QTc had a significant but relatively weak correlation with most of the T‐wave morphology parameters and QTd with all the T‐wave morphology parameters (Table 4).
Table 1.
Clinical characteristics in study patients at baseline
| Variable | Alive n = 1,744 | CD n = 95 | SCD n = 44 | NSCD n = 51 | NCD n = 106 |
|---|---|---|---|---|---|
| Age | 66.3 ± 8.5 | 71.8 ± 7.8*** | 69.6 ± 7.6* | 73.8 ± 7.6*** | 72.4 ± 7.9*** |
| Gender (%) | 67.0 | 76.8** | 79.5 | 74.5 | 81.1** |
| BMI | 28.3 ± 4.5 | 28.9 ± 5.3 | 28.7 ± 4.8 | 29.1 ± 5.7 | 28.2 ± 4.9 |
| BP, systolic | 147.3 ± 24.4 | 146.7 ± 25.6 | 146.0 ± 22.2 | 147.2 ± 28.5 | 145.8 ± 28.4 |
| BP, diastolic | 80.8 ± 11.3 | 79.3 ± 11.2 | 79.5 ± 10.8 | 79.1 ± 11.6 | 79.3 ± 14.6 |
| DM, (%) | 40.8 | 67.4*** | 68.2 *** | 66.7*** | 55.7** |
| Smoker (%) | 14.4 | 18.8 | 24.0 | 13.0 | 25.0 |
| F smoker (%) | 45.7 | 54.7 | 50.0 | 58.3 | 61.7** |
| CF smoker (%) | 50.3 | 58.9 | 56.8 | 60.8 | 66.0** |
| Alcohol | 2.0 ± 4.4 | 3.9 ± 8.3* | 5.5 ± 10.1* | 2.6 ± 6.3 | 3.1 ± 9.1 |
| NYHA class I/II/III (%) | 97/2.8/0.4 | 92/7.5/1.1*** | 88/12/0.0* | 95/3.9/2.0** | 95/4.7/0.0 |
| Claudic. (%) | 6.2 | 15.1** | 13.6 | 16.3* | 17.1*** |
| β‐blocker (%) | 87.3 | 91.6 | 88.6 | 94.1 | 89.6 |
| ACE (%) | 40.4 | 40.0 | 38.6 | 41.2 | 39.6 |
| AT2 (%) | 28.3 | 33.7 | 36.4 | 31.4 | 32.1 |
| Lipid l.m. (%) | 91.8 | 85.3* | 90.9 | 80.4** | 90.6 |
| Anti‐t.m. (%) | 97.9 | 96.8 | 100.0 | 94.1 | 100.0 |
| Diuretics (%) | 32.0 | 55.8*** | 52.3** | 58.8*** | 53.8*** |
| Ca‐blocker (%) | 24.2 | 27.4 | 25.0 | 29.4 | 24.5 |
| Nitrates (%) | 35.0 | 61.1*** | 56.8** | 64.7*** | 46.2* |
| Anti‐a.m. (%) | 0.8 | 3.2 | 0.0 | 5.9* | 0.9 |
| LVMI (g/m2) | 106.7 ± 26.7 | 124.2 ± 34.2*** | 116.9 ± 29.0* | 130.4 ± 37.3*** | 112.2 ± 25.9* |
| LVEF (%) | 64.4 ± 8.7 | 57.7 ± 15.0*** | 59.0 ± 13.8* | 56.7 ± 15.9** | 64.1 ± 10.1 |
| LVEDD (mm) | 50.3 ± 6.0 | 52.8 ± 10.1* | 51.8 ± 10.3 | 53.6 ± 10.0* | 49.2 ± 6.7 |
| LVESD (mm) | 32.1 ± 6.2 | 37.4 ± 11.1*** | 36.3 ± 11.3* | 38.4 ± 10.8*** | 32.2 ± 6.9 |
| LBBB (%) | 2.6 | 4.5 | 4.9 | 4.3 | 3.0 |
| RBBB (%) | 4.3 | 8.0 | 7.3 | 8.5 | 11.1** |
| Fam. SCD (%) | 35.9 | 35.1 | 32.6 | 37.3 | 25.7* |
The values are mean ± SD or percentages. ACE, angiotensin‐converting enzyme inhibitor; Alcohol, alcohol consumption portions/week; Anti‐a.m., antiarrhythmic medication; Anti‐t.m., antithrombotic medication; AT2, angiotensin receptor 2 inhibitor; BMI, body mass index; BP, blood pressure; CD, cardiac death; CF smoker, current/former smoker; Claudic., claudication; DM, diabetes mellitus type 2; Fam. SCD, a history of sudden cardiac death in the family; F smoker, former smoker; Gender, male gender; Lipid l.m., lipid lowering medication; LVEDD, left ventricular end‐diastolic diameter; LVEF, left ventricular ejection fraction; LVESD, left ventricular end‐systolic diameter; LVMI, left ventricular mass index; NCD, noncardiac death; Nitrates, long‐acting nitrates; NSCD, nonsudden cardiac death; NYHA, New York Heart Association class; SCD, sudden cardiac death.
*p < .05, **p < .01, ***p < .001 when compared with the subjects who remained alive.
Table 2.
T‐wave morphology parameters in study patients at baseline
| Variable | Alive n = 1,744 | CD n = 95 | SCD n = 44 | NSCD n = 51 | NCD n = 106 |
|---|---|---|---|---|---|
| Heart rate | 60.6 ± 10.2 | 64.0 ± 12.6* | 62.5 ± 11.1 | 65.3 ± 13.8* | 61.7 ± 11.2 |
| QTc | 414.8 ± 25.3 | 427.5 ± 28.0*** | 427.9 ± 27.5** | 427.1 ± 28.7** | 424.6 ± 30.5** |
| QTd | 28.1 ± 14.2 | 29.0 ± 14.1 | 28.0 ± 9.1 | 29.7 ± 17.4 | 30.8 ± 15.1 |
| TWLD | 53.7 ± 16.0 | 50.7 ± 12.6 | 48.9 ± 11.1* | 52.2 ± 13.7 | 51.4 ± 12.5 |
| T_PCA1 | 555.5 ± 366.5 | 476.8 ± 277.0* | 501.5 ± 299.6 | 455.2 ± 257.1 | 515.8 ± 308.9 |
| T_PCA2 | 24.7 ± 16.3 | 28.4 ± 19.3 | 26.0 ± 15.7 | 30.5 ± 21.8 | 26.2 ± 17.1 |
| T_PCA3 | 8.1 ± 6.6 | 8.5 ± 6.0 | 7.2 ± 3.3 | 9.7 ± 7.5 | 8.8 ± 7.9 |
| TMD | 34.8 ± 27.1 | 51.4 ± 29.1*** | 44.3 ± 28.8* | 57.6 ± 28.3*** | 37.4 ± 28.5 |
| TCRT | 0.26 ± 0.57 | −0.10 ± 0.63*** | 0.03 ± 0.69* | −0.22 ± 0.56*** | 0.09 ± 0.61** |
| QRSTmean | 89.1 ± 39.5 | 110.0 ± 42.4*** | 100.8 ± 47.4 | 118.0 ± 36.2*** | 101.7 ± 38.7** |
| TWAD | 0.35 ± 0.44 | 0.06 ± 0.50*** | 0.22 ± 0.51 | ‐0.08 ± 0.44*** | 0.30 ± 0.44 |
| TLL3D | 4.4 ± 1.1 | 4.6 ± 1.3 | 4.7 ± 1.2 | 4.6 ± 1.4 | 4.5 ± 1.3 |
QTc, QT interval corrected for heart rate by the Bazett's formula; QTd, QT dispersion calculated by subtracting the shortest QT interval from the longest among the 12 leads of ECG; TWLD, T‐wave loop dispersion; T_PCA1‐3, T‐wave loop principal component analyses 1‐3; TMD, T‐wave morphology dispersion; TCRT, total cosine R‐to‐T; QRSTmean, mean angle between QRS loop and T‐wave loop vectors; TWAD, T‐wave area dispersion; TLL3D, T‐loop length measured in 3D.
*p < .05, **p < .01, ***p < .001 when compared with the subjects who remained alive.
Table 3.
Univariate and multivariate predictors of cardiac death, sudden cardiac death, nonsudden cardiac death, and noncardiac death
| Variable | CD | SCD | NSCD | NCD | |||||
|---|---|---|---|---|---|---|---|---|---|
| HR | 95% CI | HR | 95% CI | HR | 95% CI | HR | 95% CI | ||
| Age | uv | 1.093 | 1.062–1.124*** | 1.052 | 1.011–1.093* | 1.140 | 1.093–1.190*** | 1.105 | 1.075–1.136*** |
| mv | 1.104 | 1.070–1.138*** | 1.071 | 1.027–1.117** | 1.136 | 1.087–1.187*** | 1.114 | 1.080–1.149*** | |
| Gender | uv | 1.643 | 1.020–2.647* | 1.939 | 0.932–4.034 | 1.458 | 0.777–2.738 | 2.120 | 1.303–3.449** |
| mv | 2.200 | 1.244–3.890** | |||||||
| DM | uv | 2.742 | 1.784–4.215*** | 2.850 | 1.506–5.390** | 2.743 | 1.532–4.913** | 1.683 | 1.144–2.474** |
| mv | 2.124 | 1.355–3.330** | 2.288 | 1.162–4.504* | 2.212 | 1.220–4.011** | |||
| F smoker | uv | 1.963 | 1.295–2.975** | ||||||
| mv | 1.745 | 1.095–2.781* | |||||||
| Alcohol | uv | 1.058 | 1.030–1.087*** | 1.080 | 1.047–1.114*** | ||||
| mv | 1.094 | 1.065–1.123*** | 1.103 | 1.070–1.138*** | |||||
| NYHA | uv | 3.337 | 2.109–5.281*** | 2.743 | 1.425–5.281** | 4.074 | 2.131–7.786*** | ||
| mv | 1.729 | 1.056–2.833* | |||||||
| Lipid l.m. | uv | 0.548 | 0.311–0.966* | 0.385 | 0.193–0.769** | ||||
| mv | 0.421 | 0.200–0.887* | |||||||
| Diuretic drugs | uv | 2.540 | 1.694–3.808*** | 2.239 | 1.239–4.047** | 2.921 | 1.672–5.100*** | 2.337 | 1.595–3.424*** |
| mv | 1.697 | 1.115–2.583* | |||||||
| Nitrates | uv | 2.696 | 1.785–4.072*** | 2.312 | 1.273–4.199** | 3.196 | 1.800–5.676*** | 1.519 | 1.037–2.226* |
| mv | 1.833 | 1.181–2.846** | 2.050 | 1.088–3.863* | 2.211 | 1.210–4.040* | |||
| Claudication | uv | 2.288 | 1.296–4.041** | 2.505 | 1.174–5.345* | 2.634 | 1.585–4.377*** | ||
| mv | 1.879 | 1.033–3.416* | |||||||
| Anti‐a.m. | uv | 5.371 | 1.672–17.26** | ||||||
| mv | |||||||||
| LVMI | uv | 1.019 | 1.013–1.024*** | 1.013 | 1.004–1.022** | 1.023 | 1.016–1.030*** | 1.008 | 1.002–1.015* |
| mv | 1.014 | 1.005–1.023** | |||||||
| LVEF | uv | 0.941 | 0.926–0.956*** | 0.944 | 0.920–0.969*** | 0.933 | 0.914–0.953*** | ||
| mv | 0.963 | 0.940–0.986** | 0.942 | 0.917–0.967*** | 0.948 | 0.925–0.972*** | |||
| LVEDD | uv | 1.060 | 1.029–1.093*** | 1.080 | 1.038–1.124*** | ||||
| mv | |||||||||
| LVESD | uv | 1.086 | 1.063–1.110*** | 1.076 | 1.039–1.114*** | 1.100 | 1.070–1.132*** | ||
| mv | 1.039 | 1.008–1.071* | |||||||
| Fam. SCD | uv | 0.642 | 0.414–0.994* | ||||||
| mv | 0.536 | 0.331–0.867* | |||||||
| QTc | uv | 1.018 | 1.010–1.026*** | 1.019 | 1.008–1.030** | 1.018 | 1.007–1.028** | 1.014 | 1.007–1.021*** |
| mv | 1.006 | 0.998–1.014 | 1.010 | 0.999–1.022 | 1.003 | 0.993–1.013 | 1.011 | 1.003–1.018** | |
| TWLD | uv | 0.974 | 0.949–1.0002 | ||||||
| mv | 0.980 | 0.954–1.006 | |||||||
| T_PCA1 | uv | 0.999 | 0.999–0.99995* | ||||||
| mv | 1.000 | 0.999–1.001 | |||||||
| TMD | uv | 1.020 | 1.013–1.028*** | 1.013 | 1.002–1.023* | 1.028 | 1.018–1.038*** | ||
| mv | 1.015 | 1.007–1.023*** | 1.008 | 0.996–1.020 | 1.017 | 1.006–1.029** | |||
| TCRT | uv | 0.377 | 0.280–0.530*** | 0.526 | 0.320–0.866* | 0.271 | 0.168–0.439*** | 0.620 | 0.449–0.857** |
| mv | 0.598 | 0.412–0.866** | 0.803 | 0.466–1.385 | 0.546 | 0.333–0.893* | 0.848 | 0.597–1.205 | |
| QRSTmean | uv | 1.013 | 1.008–1.019*** | 1.019 | 1.011–1.027*** | 1.008 | 1.003–1.013** | ||
| mv | 1.004 | 0.998–1.010 | 1.007 | 0.999–1.015 | 1.003 | 0.998–1.009 | |||
| TWAD | uv | 0.306 | 0.206–0.453*** | 0.185 | 0.107–0.317*** | ||||
| mv | 0.474 | 0.305–0.737*** | 0.365 | 0.197–0.678** | |||||
CI, confidence interval; HR, hazards ratio; mv, multivariate; uv, univariate. Other abbreviations are same as in Tables 1 and 2. Only the variables that differed significantly in univariate comparisons in Tables 1 and 2 were included in the Cox univariate regression analysis. Only the HRs that remained significant in the multivariate Cox regression analysis are shown for the clinical variables, whereas all the multivariate HRs are show for T‐wave morphology parameters whether significant or not.
*p < .05, **p < .01, ***p < .001.
Table 4.
Correlation between standard QT measurements and T‐wave morphology parameters
| TWLD | T_PCA1 | T_PCA2 | T_PCA3 | TMD | TCRT | QRSTmean | TWAD | TLL3D | |
|---|---|---|---|---|---|---|---|---|---|
| QTc | −0.003 (p = .91) | −0.160 (p < .001) | 0.154 (p < .001) | 0.096 (p < .001) | 0.208 (p < .001) | −0.203 (p < .001) | 0.220 (p < .001) | −0.177 (p < .001) | 0.079 (p = .001) |
| QTd | −0.151 (p < .001) | −0.125 (p < .001) | 0.136 (p < .001) | 0.051 (p = .025) | 0.136 (p < .001) | −0.184 (p < .001) | 0.177 (p < .001) | −0.147 (p < .001) | 0.158 (p < .001) |
The values are the Pearson correlation coefficients, (p‐values in parentheses). The abbreviations are the same as in the Table 2.
3.2. Discrimination and reclassification accuracy of the T‐wave morphology parameters
When including TMD, TCRT, and TWAD to the clinical risk model for CD, the C‐index increased from 0.810 to 0.823 improving IDI significantly, and adding these parameters separately or together in the clinical prediction model improved NRI significantly (Table 5).
Table 5.
Discrimination and reclassification accuracy of risk markers of cardiac death
| C‐index | 95% CI | NRI | 95% CI | p‐value | IDI | 95% CI | p‐value | |
|---|---|---|---|---|---|---|---|---|
| Clinical model | 0.810 | 0.765–0.856 | ‐ | ‐ | ‐ | ‐ | ‐ | ‐ |
| +TMD | 0.819 | 0.775–0.864 | 0.350 | 0.136–0.563 | .00132 | 0.0109 | 0.0023–0.0195 | .013 |
| +TCRT | 0.820 | 0.777–0.863 | 0.430 | 0.222–0.639 | .00005 | 0.0057 | −0.0010 to 0.0125 | .097 |
| +TWAD | 0.821 | 0.778–0.865 | 0.380 | 0.167–0.594 | .00048 | 0.0068 | −0.0009 to 0.0145 | .083 |
| +TMD, TCRT and TWAD | 0.823 | 0.780–0.867 | 0.257 | 0.043–0.471 | .01858 | 0.0118 | 0.0028–0.0208 | .010 |
CI, confidence interval; IDI, integrated discrimination index; NRI, net reclassification index; TCRT, total cosine R‐to‐T; TMD, T‐wave morphology dispersion; TWAD, T‐wave area dispersion. The clinical risk model for cardiac death included: age, diabetes mellitus, alcohol consumption, New York Heart Association class, lipid lowering medication, long‐acting nitrate medication, left ventricular ejection fraction and left ventricular size.
3.3. Cumulative proportional probabilities of different modes of death in relation to T‐wave area dispersion
Cumulative proportional probabilities of CD, SCD, NSCD, and NCD in relation to TWAD are shown in Figure 1. The optimal cut point for CD obtained from ROC curve was used in the analyses.
Figure 1.
Cumulative proportional probability of cardiac death (top left), sudden cardiac death (top right), nonsudden cardiac death (bottom left), and noncardiac death (bottom right) in patients with T‐wave area dispersion (TWAD) < or ≥−0.06. The cutpoint was optimized from the receiver operating characteristics (ROC) curve for cardiac death
4. DISCUSSION
The main finding of the present study was that several T‐wave morphology parameters were associated with the risk of CD in patients with CAD in the current treatment era. After adjustments with relevant risk markers, higher values of TMD, and lower values of TWAD and TCRT still retained their predictive power for the occurrence of CD. Including these parameters improved the discrimination accuracy of the clinical risk model for CD. Some parameters, such as TWLD, TMD, and TCRT were associated with SCD in univariate comparisons. However, after relevant adjustment in the multivariate model, the association of theses parameters with SCD did not remain significant. NSCD was associated with TMD, TCRT, and TWAD, even after relevant adjustments. The association of T‐wave morphology parameters with CD was more driven by their association with NSCD than SCD. TCRT and QRSTmean were also associated with NCD in univariate comparisons but not after multivariate adjustments. QTc was associated with all modes of death in univariate analyses; however, it retained significant association only with NCD after relevant adjustments. The correlation of QTc with the T‐wave morphology parameters, although significant, was relatively weak. Taken together, the T‐wave morphology parameters yield additional information on the risk of CD in patients with CAD compared to QTc.
One of the predictors of CD in the present study was TMD. It describes the spatial heterogeneity of T‐wave morphology with higher values suggesting inhomogeneous repolarization (Porthan et al., 2013). In a previous study of 280 consecutive postinfarction patients, who were followed up 32 ± 10 months on average, TMD did not have significant association with primary combined endpoint of all‐cause mortality, sustained ventricular tachycardia (VT) or resuscitated ventricular fibrillation (VF), or with secondary endpoint of arrhythmic events (Zabel et al., 2000). In another study in 813 male US veterans with cardiovascular disease TMD exhibited univariate but not independent predictive value for total mortality (Zabel et al., 2002). In a study by Porthan et al., (2013) in general population including 5,618 subjects, who were followed up 7.7 ± 1.4 years on average, TMD was associated with risk of SCD and NSCD even after multivariate adjustments. In that study, there were 72 deaths that were classified as SCD compared with only 44 SCD cases in the present study in patients with CAD which obviously has weakened the statistical power of T‐morphology parameters in prediction of SCD in our study.
In the present study, TCRT also was an independent predictor of CD. It has been suggested that TCRT may be a marker of global repolarization heterogeneity or ventricular electrical dyssynchrony (Perkiömäki et al., 2006; Smetana et al., 2002). In the above‐mentioned study in postinfarction patients by Zabel et al., lower values of TCRT independently predicted the primary endpoint of all‐cause mortality, sustained VT, and resuscitated VF, and tended to independently predict the secondary endpoint of arrhythmic events (Zabel et al., 2000). In another study including 437 postinfarction patients, TCRT was a univariate predictor of cardiac mortality during an average follow‐up period of 43 ± 14 months but lost its predictive power after relevant multivariate adjustments (Perkiömäki et al., 2006). Attenuated postexercise recovery of TCRT has been shown to be an independent predictor of CD and SCD (Kenttä et al., 2012), an observation, which emphasizes the possibility to gain more prognostic information by evaluating spatiotemporal, instead of just spatial, heterogeneity from ECG, e.g., from exercise ECGs or Holter recordings. In the above‐mentioned study including US veterans, TCRT had a univariate association with the risk of total mortality (Zabel et al., 2002). Aro et al. found in their study including 10,957 middle‐aged subjects from general population, who were followed up for 30 ± 11 years on average that frontal QRS‐T angle (a substitute for TCRT) predicted sudden arrhythmic death (Aro et al., 2012). In the present analysis, QRSTmean was a predictor of CD, NSCD and NCD on univariate analyses but did not have significant association with SCD. Porthan et al. observed in their study in general population that TCRT remained as a predictor of SCD even after adjustments in multivariate model (Porthan et al., 2013).
Preliminary results in a general population have suggested that lower values of TWAD, a measure of repolarization heterogeneity, are an independent predictor of SCD (Kenttä et al., 2015). In our present study in patients with CAD, TWAD was a powerful independent predictor of CD and NSCD but did not have significant association with SCD as a continuous variable in the group comparison. However, when optimized cut point was used, TWAD as a dichotomized parameter predicted the proportional probability of SCD during follow‐up.
TWLD, which describes the shape of T‐wave loop and hence repolarization heterogeneity, has been shown to be independently associated with cardiac death and arrhythmic events in postinfarction patients (Perkiömäki et al., 2006; Zabel et al., 2000). However, in the study of male US veterans with cardiovascular disease, TWLD did not predict total mortality (Zabel et al., 2002). In the present study, lower values of TWLD had a univariate association with SCD but not with other modes of death. Newer methods to describe electrical heterogeneity from ECG, such as second central moment analysis, have continuously been developed. T‐wave heterogeneity based on this methodology has already been shown to be an independent predictor of SCD (Kenttä et al., 2016).
All the T‐wave morphology parameters, which were found to yield prognostic information in the present analysis, can be considered to describe the heterogeneity of repolarization in myocardium. Heterogeneity of repolarization is increased in pathological states and reflected to T‐wave morphology in complex way originating from ionic currents at cellular level, and transformed in endo‐, midmyo‐, and epicardium (Antzelevitch, 2007; Boukens, Walton, Meijborg, & Coronel, 2016). Heterogeneity of ventricular repolarization increases the risk of reentrant life‐threatening ventricular tachyarrhythmias (Yan et al., 2004). This is an obvious mechanistic link between abnormal values of T‐wave morphology parameters and arrhythmic cardiac death. As described above, in some studies T‐morphology parameters have been shown to predict SCD mainly in general populations (Aro et al., 2012; Kenttä et al., 2012, 2015; Porthan et al., 2013). In our present study in patients with CAD treated according to contemporary guidelines, T‐wave morphology parameters were more closely associated with NSCD than SCD. Particularly, in cases of NSCD, the abnormalities in T‐wave morphology parameters obviously are a forerunner of gradually worsening myocardial pathology. In addition, in the final phase before NSCD ventricular arrhythmias often hasten the death. SCD in patients with CAD is often linked with abrupt plaque rupture/complication which is initiated by factors other than repolarization heterogeneity. This is one potential explanation why the parameters describing repolarization heterogeneity were better predictors of NSCD than SCD. However, the exact mechanisms by which T‐wave morphology disturbances were more closely associated with NSCD than SCD remain somewhat elusive and should be assessed in further studies.
There are some limitations in the present study. We did not do repeated measurements of T‐wave morphology during the follow‐up, instead the analyses were based on baseline values. The high risk patients, such as those with planned or existing implantable‐cardioverter defibrillator, were excluded from the study. This kind of patients with high risk of SCD would more likely have repolarization heterogeneity. However, there were only few patients who were excluded on this basis.
In conclusion, T‐wave morphology descriptors of repolarization heterogeneity predict CD in patients with CAD in current treatment era and improve the discrimination accuracy of the clinical risk model for CD.
CONFLICT OF INTEREST
The authors declare that they have no conflicts of interest.
Pirkola JM, Konttinen M, Kenttä TV, et al. Prognostic value of T‐wave morphology parameters in coronary artery disease in current treatment era. Ann Noninvasive Electrocardiol. 2018;23:e12539 10.1111/anec.12539
Funding Information
The study is supported by a grant from the Sigrid Juselius Foundation, Helsinki, Finland, and the Finnish Foundation for Cardiovascular Research, Helsinki, Finland.
REFERENCES
- Acar, B. , Yi, G. , Hnatkova, K. , & Malik, M. (1999). Spatial, temporal and wavefront direction characteristics of 12‐lead T‐wave morphology. Medical and Biological Engineering and Computing, 37, 574–584. 10.1007/BF02513351 [DOI] [PubMed] [Google Scholar]
- Adabag, A. S. , Luepker, R. V. , Roger, V. L. , & Gersh, B. J. (2010). Sudden cardiac death: Epidemiology and risk factors. National Review of Cardiology, 7, 216–225. 10.1038/nrcardio.2010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Antzelevitch, C. (2007). Heterogeneity and cardiac arrhythmias: An overview. Heart Rhythm, 4, 964–972. 10.1016/j.hrthm.2007.03.036 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Aro, A. L. , Huikuri, H. V. , Tikkanen, J. T. , Junttila, M. J. , Rissanen, H. A. , Reunanen, A. , & Anttonen, O. (2012). QRS‐T angle as a predictor of sudden cardiac death in a middle‐aged general population. Europace, 14, 872–876. 10.1093/europace/eur393 [DOI] [PubMed] [Google Scholar]
- Boukens, B. J. , Walton, R. , Meijborg, V. M. , & Coronel, R. (2016). Transmural electrophysiological heterogeneity, the T‐wave and ventricular arrhythmias. Progress in Biophysics and Molecular Biology, 122, 202–214. 10.1016/j.pbiomolbio.2016.05.009 [DOI] [PubMed] [Google Scholar]
- Deo, R. , & Albert, C. M. (2012). Epidemiology and genetics of sudden cardiac death. Circulation, 125, 620–637. 10.1161/CIRCULATIONAHA.111.023838 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Engdahl, J. , Holmberg, M. , Karlson, B. , Luepker, R. , & Herlitz, J. (2002). The epidemiology of out‐of‐hospital ‘sudden’ cardiac arrest. Resuscitation, 52, 235–245. 10.1016/S0300-9572(01)00464-6 [DOI] [PubMed] [Google Scholar]
- Goldberger, J. J. , Buxton, A. E. , Cain, M. , Constantini, O. , Exner, D. V. , Knight, B. P. , … Zipes, D. P. (2011). Risk stratification for arrhtyhmic sudden cardiac death: Identifying the roadblocks. Circulation, 123, 2423–2430. 10.1161/CIRCULATIONAHA.110.959734 [DOI] [PubMed] [Google Scholar]
- Kardys, I. , Kors, J. A. , van der Meer, I. M. , Hofman, A. , van der Kuip, D. A. , & Witteman, J. C. (2003). Spatial QRS‐T angle predicts cardiac death in a general population. European Heart Journal, 24, 1357–1364. 10.1016/S0195-668X(03)00203-3 [DOI] [PubMed] [Google Scholar]
- Kenttä, T. , Nearing, B. , Porthan, K. , Tikkanen, J. T. , Viitasalo, M. , & Nieminen, M. S. (2015). Dispersion of T‐wave area in left precordial leads predicts sudden cardiac death in standard 12‐lead electrocardiogram. Heart Rhythm, 12, S209. [Google Scholar]
- Kenttä, T. V. , Nearing, B. D. , Porthan, K. , Tikkanen, J. T. , Viitasalo, M. , Nieminen, M. S. , … Verrier, R. L. (2016). Prediction of sudden cardiac death with automated high‐throughput analysis of heterogeneity in standard resting 12‐lead electrocardiograms. Heart Rhythm, 13, 713–720. 10.1016/j.hrthm.2015.11.035 [DOI] [PubMed] [Google Scholar]
- Kenttä, T. , Viik, J. , & Huikuri, H. (2012). Postexercise recovery of the spatial QRS/T angle as the predictor of sudden cardiac death. Heart Rhythm, 9, 1083–1089. 10.1016/j.hrthm.2012.02.030 [DOI] [PubMed] [Google Scholar]
- Kors, J. A. , de Bruyne, M. C. , Hoes, A. W. , van Herpen, G. , Hofman, A. , van Bemmel, J. H. , & Grobbee, D. E. (1998). T axis as an indicator of risk of cardiac events in elderly people. Lancet, 352, 601–605. 10.1016/S0140-6736(97)10190-8 [DOI] [PubMed] [Google Scholar]
- Lang, R. M. , Badano, L. P. , Mor‐Avi, V. , Afilalo, J. , Armstrong, A. , Ernande, L. , … Voigt, J. U. (2015). Recommendations for cardiac chamber quantification by echocardiography in adults: An update from the American society of echocardiography and the European association of cardiovascular imaging. European Heart Journal of Cardiovascular Imaging, 16, 233–270. 10.1093/ehjci/jev014 [DOI] [PubMed] [Google Scholar]
- Perkiömäki, J. S. , Hyytinen‐Oinas, M. , Karsikas, M. , Seppänen, T. , Hnatkova, K. , Malik, M. , & Huikuri, H. V. (2006). Usefulness of T‐wave loop and QRS complex loop to predict mortality after acute myocardial infarction. American Journal of Cardiology, 97, 353–360. 10.1016/j.amjcard.2005.08.052 [DOI] [PubMed] [Google Scholar]
- Piironen, M. , Ukkola, O. , Huikuri, H. , Havulinna, A. S. , Koukkunen, H. , Mustonen, J. , … Salomaa, V. (2017). Trends in long‐term prognosis after acute coronary syndrome. European Journal of Preventive Cardiology, 24, 274–280. 10.1177/2047487316679522 [DOI] [PubMed] [Google Scholar]
- Porthan, K. , Viitasalo, M. , Toivonen, L. , Havulinna, A. S. , Jula, A. , & Tikkanen, J. T. (2013). Predictive value of electrocardiographic T‐wave morphology parameters and T‐wave peak to T‐wave end interval for sudden cardiac death in the general population. Circulation: Arrhythmia and Electrophysiology, 6, 690–696. 10.1161/CIRCEP.113.000356 [DOI] [PubMed] [Google Scholar]
- Smetana, P. , Batchvarov, V. N. , Hnatkova, K. , Camm, A. J. , & Malik, M. (2002). Sex differences in repolarization homogeneity and its circadian pattern. American Journal of Physiology. Heart and Circulatory Physiology, 282, H1889–H1897. 10.1152/ajpheart.00962.2001 [DOI] [PubMed] [Google Scholar]
- Verrier, R. L. , & Huikuri, H. V. (2017). Tracking interlead heterogeneity of R‐ and T‐wave morphology to disclose latent risk for sudden cardiac death. Heart Rhythm, 14, 1466–1475. 10.1016/j.hrthm.2017.06.017 [DOI] [PubMed] [Google Scholar]
- WHO . (2006). Definition and diagnosis of diabetes mellitus and intermediate hyperglycaemia. Report of a WHO/IDF consultation. ISBN: 978 92 4 159493 6.
- Yan, G. X. , Joshi, A. , Guo, D. , Hlaing, T. , Martin, J. , Xu, X. , & Kowey, P. R. (2004). Phase 2 reentry as a trigger to initiate ventricular fibrillation during early acute myocardial ischemia. Circulation, 110, 1036–1041. 10.1161/01.CIR.0000140258.09964.19 [DOI] [PubMed] [Google Scholar]
- Zabel, M. , Acar, B. , Klingenheben, T. , Franz, M. R. , Hohnloser, S. H. , & Malik, M. (2000). Analysis of 12‐lead T‐wave morphology for risk stratification after myocardial infarction. Circulation, 102, 1252–1257. 10.1161/01.CIR.102.11.1252 [DOI] [PubMed] [Google Scholar]
- Zabel, M. , Malik, M. , Hnatkova, K. , Papademetriou, V. , Pittaras, A. , Fletcher, R. D. , & Franz, M. R. (2002). Analysis of T‐wave morphology from the 12‐lead electrocardiogram for prediction of long‐term prognosis in male US veterans. Circulation, 105, 1066–1070. 10.1161/hc0902.104598 [DOI] [PubMed] [Google Scholar]
- Zipes, D. , & Wellens, H. J. J. (1998). Sudden Cardiac Death. Circulation, 98, 2334–2351. 10.1161/01.CIR.98.21.2334 [DOI] [PubMed] [Google Scholar]