Electrocardiographic lead aVR is often ignored in clinical practice. The aim of this study was to investigate whether ST-T wave amplitude in lead aVR predicts cardiovascular (CV) mortality and if this variable adds value to a traditional risk prediction model. A total of 7,928 participants enrolled in the National Health and Nutrition Examination Survey (NHANES) III with electrocardiographic data available were included. Each participant had 13.5 ± 3.8 years of follow-up. The study sample was stratified according to ST-segment amplitude and T-wave amplitude in lead aVR. ST-segment elevation (>8 μV) in lead aVR was predictive of CV mortality in the multivariate analysis when not accounting for T-wave amplitude. The finding lost significance after including T-wave amplitude in the model. A positive T wave in lead aVR (>0 mV) was the strongest multivariate predictor of CV mortality (hazard ratio 3.37, p <0.01). The addition of T-wave amplitude in lead aVR to the Framingham risk score led to a net reclassification improvement of 2.7% of subjects with CV events and 2.3% of subjects with no events (p <0.01). Furthermore, in the intermediate-risk category, 20.0% of the subjects in the CV event group and 9.1% of subjects in the no-event group were appropriately reclassified. The absolute integrated discrimination improvement was 0.012 (p <0.01), and the relative integrated discrimination improvement was 11%. In conclusion, T-wave amplitude in lead aVR independently predicts CV mortality in a cross-sectional United States population. Adding T-wave abnormalities in lead aVR to the Framingham risk score improves model discrimination and calibration with better reclassification of intermediate-risk subjects.
Electrocardiographic lead aVR, an augmented unipolar limb lead with the positive electrode on the right arm, provides important diagnostic information regarding coronary ischemia and myocardial injury, pulmonary embolism, and rhythm disorders. However, lead aVR is often overlooked in clinical practice. There is evidence to also suggest its role as a prognostic marker for cardiovascular (CV) mortality. The current research evaluating the prognostic significance of ST-T wave amplitude in lead aVR is based on studies either restricted to men who underwent electrocardiography for clinical indications or that were conducted outside the United States. The Framingham risk score (FRS) classifies a substantial portion of the United States population as being at intermediate risk (FRS 5% to 20%) on the basis of traditional risk factors. CV risk factor management in this population is not well delineated. To achieve more efficient risk stratification, and to reduce the CV mortality burden, we believe that there currently exists a need for a low-cost, easily available diagnostic tool for reclassifying intermediate-risk subjects into either higher or lower risk strata. Electrocardiography is a widely available, inexpensive, and noninvasive tool that has been shown to have utility for refining the FRS model. Despite this, the United States Preventive Services Task Force does not currently support the use of electrocardiography for screening purposes, because of a lack of evidence. We therefore sought to assess the relation between ST-T wave abnormalities in aVR and CV mortality by performing a retrospective observational cohort study using prospectively collected data from the National Health and Nutrition Examination Survey (NHANES) III (1988 to 1994). We also assessed the additive value of including lead aVR changes in the existing risk prediction models for CV mortality.
Methods
NHANES III, conducted by the National Center for Health Statistics, includes data from oral surveys and general health examinations. It was designed to assess the demographic, socioeconomic, dietary, and overall health status of a nationally representative sample of noninstitutionalized patients from all 50 states. The study population consisted of 39,695 subjects aged >40 years selected over the course of 6 years. Of those subjects selected to participate, 30,818 (78%) completed the health examination.
Our initial cohort of 8,561 subjects was selected from all adults enrolled in NHANES III from 1988 to 1994 with available electrocardiographic data. We then excluded patients with missing data on mortality (n = 4), missing data on ST-T wave abnormalities (n = 58), QRS duration >120 ms (n = 570), or atrial fibrillation (n = 1). Our final study population included a total of 7,928 participants, with a mean follow-up period of 13.5 ± 3.8 years per patient, giving us a total follow-up period of 107,028 person-years.
In NHANES III, each of the participants aged >40 years underwent baseline rest 12-lead electrocardiography in a supine position at a mobile examination center. A Marquette MAC 12 (Marquette Medical Systems, Inc., Milwaukee, Wisconsin) was used to record the electrocardiogram. The system uses multiple leads simultaneously to improve the quality of the recording. On the basis of predefined standards, the mobile examination center physician was responsible for approving the quality of the tracings. Electrocardiographic analysis was performed at the Epidemiological Cardiology Research Center at Wake Forest University. The Novacode electrocardiographic program was used for the analysis, which classified electrocardiograms per the Minnesota Code Classification System.
The primary outcome considered in this analysis was CV mortality. The NHANES III linked mortality public-use file was used to obtain mortality status. It provided mortality data through December 31, 2006. Matching records to the National Death Index was used as the primary determinant of mortality for eligible NHANES III participants. Underlying causes of death were provided by death certificate data contained in the same mortality files and were classified according to the International Classification of Diseases, 10th Revision, injury and cause-of-death guidelines. Codes for CV disease were identified according to American Heart Association guidelines.
NHANES III had a complex nonrandom, multistage, stratified sample design. All analyses were performed using the designated weighting, which was specified in the NHANES III data set to minimize bias. The total NHANES III pseudostratum was used as our stratum variable. The total NHANES III pseudoprimary sampling unit was used as our survey sampling unit and the total mobile examination center final weight as our sampling unit weight. For categorical variables, chi-square analysis was used to evaluate group differences. For continuous variables, to evaluate group differences, 1-way analysis of variance was used if the variable was normally distributed, and the Kruskal-Wallis test was used if the distribution was not normal. Cox proportional-hazards regression modeling was used to calculate the hazard ratio of CV mortality.
For the multivariate analysis, ST-segment amplitudes in lead aVR were split into quartiles, because they were not normally distributed. T waves in lead aVR were categorized by amplitude (<−2, −2 to −1, −1 to 0, and >0 mV), as in previous studies, and for the clinical applicability of the study. Other variables entered in the regression model were age, gender, race, body mass index, hypertension, smoking status, serum cholesterol, glomerular filtration rate, serum potassium, and left ventricular mass index. Three separate multivariate models were run, accounting for ST-segment amplitude, T-wave amplitude, and ST-segment and T-wave amplitude added together in the same model.
For reclassification, we compared 2 different predictive models. Model A included the FRS only, which includes age, gender, systolic blood pressure, smoking history, serum cholesterol level, and serum high-density lipoprotein level. Model B included the FRS and the variable T-wave amplitude. The differences in global measure of model fit were assessed using a likelihood ratio test. The Bayesian information criterion (for which lower values imply better fit) was calculated to evaluate the improvement in the global measure of model fit after the addition of T-wave amplitude to model A. The Bayesian information criterion is a likelihood-based measure that takes into account the number of variables in a predictive model and prefers a model with fewer variables if it provides equally good prediction. The C-statistic was used to test the predictive ability of the new model to discriminate between patients who developed CV-related events and those who did not. C-statistics for these 2 models were compared using bootstrap sampling. To test the ability of the new model to predict the risk for future events (calibration), we calculated the Hosmer-Lemeshow statistic (a chi-square value >20 implies a lack of calibration).
The predicted 10-year risk estimate was calculated on the basis of the 2 models for all study participants with available data, and a direct comparison was made with the actual observed risk during the follow-up period. A score of 1 was assigned to every correct reclassification, which implied that every subjects who experienced an event was upgraded in the risk category, and every subject who did not experience an event was appropriately downgraded in the risk category. A score of −1 was assigned to every incorrect reclassification on the basis of the aforementioned criteria. A score of 0 was assigned to every “nonreclassification.” The integrated discrimination improvement was calculated, which was defined as the difference in the integral of sensitivity and 1−specificity over all possible cut-off values. This was the measure of improvement imparted by the variable T-wave amplitude to the FRS model. All statistical analyses were performed using STATA SE version 11.1 (StataCorp LP, College Station, Texas). Two-sided p values <0.05 were considered statistically significant.
Results
ST-segment elevation (>8 μV) in lead aVR was present in 24.2% of subjects. The prevalence of T-wave amplitude >−0.1 mV was 18.4%. Subjects with T-wave amplitude in lead aVR >−0.1 mV were older and more likely to be male, to have hypertension, to have diabetes, and to have a history of coronary artery disease ( Tables 1 and 2 ). More than 80% of participants with ST-segment amplitude >8 μV also had associated T-wave amplitude >−0.1 mV.
| Variable | All (n = 7,928) | T-Wave Amplitude in Lead aVR (mV) | p Value | |||
|---|---|---|---|---|---|---|
| <−0.2 (n = 3,048) | −0.2 to −0.1 (n = 3,417) | −0.1 to 0 (n = 1,288) | >0 (n = 175) | |||
| Age (yrs) | 59.9 ± 13.4 | 56.9 ± 13.2 | 60.5 ± 13.3 | 64.5 ± 12.6 | 63.1 ± 12.7 | <0.01 |
| Men | 45.2% | 47.5% | 42.1% | 46.6% | 61.9% | <0.01 |
| Black race | 9.2% | 7.7% | 8.8% | 13.2% | 31.1% | <0.01 |
| Body mass index (kg/m 2 ) | 27.6 ± 5.5 | 27.1 ± 5.3 | 27.6 ± 5.5 | 28.5 ± 5.6 | 28.4 ± 5.8 | <0.01 |
| Hypertension | 43.8% | 34.7% | 46.1% | 62.7% | 73.8% | <0.01 |
| Smoking | 23.1% | 23.0% | 21.9% | 25.8% | 33.3% | 0.11 |
| Hypercholesterolemia | 59.9% | 56.8% | 61.4% | 65.7% | 55.1% | 0.003 |
| Coronary artery disease | 9.8% | 5.7% | 9.9% | 19.2% | 40.0% | <0.01 |
| Myocardial infarction | 5.4% | 2.5% | 4.9% | 13.6% | 31.3% | <0.01 |
| Family history of heart attack | 11.1% | 9.6% | 12.1% | 12.6% | 11.1% | 0.06 |
| Heart failure | 2.8% | 1.3% | 2.5% | 7.4% | 19.7% | <0.01 |
| Stroke | 2.9% | 1.7% | 2.9% | 6.4% | 8.0% | <0.01 |
| Diabetes | 10.9% | 7.7% | 11.4% | 18.5% | 24.8% | <0.01 |
| Serum high-density lipoprotein (mg/dl) | 51.0 ± 16.3 | 52.1 ± 16.2 | 51.1 ± 16.3 | 48.2 ± 16.0 | 48.2 ± 15.5 | <0.01 |
| Serum cholesterol (mg/dl) | 218.3 ± 44.1 | 214.8 ± 42.6 | 219.4 ± 44.5 | 224.5 ± 45.7 | 214.1 ± 43.3 | <0.01 |
| Serum triglycerides (mg/dl) | 161.9 ± 130.3 | 151.1 ± 110.4 | 164.1 ± 145.2 | 180.1 ± 130.5 | 173.7 ± 125.0 | <0.01 |
| Serum low-density lipoprotein (mg/dl) | 136.4 ± 38.6 | 133.8 ± 38.0 | 137.2 ± 38.3 | 141.3 ± 39.8 | 132.6 ± 40.3 | <0.01 |
| Glomerular filtration rate (ml/min) | 68.2 ± 15.6 | 70.2 ± 15.2 | 67.9 ± 14.9 | 64.8 ± 17.1 | 64.4 ± 18.7 | <0.01 |
| Serum calcium (mmol/L) | 1.23 ± 0.05 | 1.23 ± 0.05 | 1.23 ± 0.05 | 1.23 ± 0.05 | 1.23 ± 0.05 | 0.43 |
| Serum potassium (mEq/L) | 4.07 ± 0.3 | 4.09 ± 0.3 | 4.05 ± 0.3 | 4.05 ± 0.4 | 4.10 ± 0.4 | 0.0001 |
| QRS duration (ms) | 96.0 ± 10.0 | 96.4 ± 9.8 | 95.4 ± 10.0 | 96.9 ± 10.4 | 96.9 ± 11.2 | <0.01 |
| Left ventricular hypertrophy | 7.5% | 2.2% | 7.1% | 22.5% | 42.1% | <0.01 |
| Variable | All (n = 7,928) | ST-Wave Changes in Lead aVR (mV) | p Value | |||
|---|---|---|---|---|---|---|
| −0.202 to −0.02 (n = 2,024) | −0.019 to −0.006 (n = 2,057) | −0.005 to 0.007 (n = 1,928) | 0.008 to 0.202 (n = 1,919) | |||
| Age (yrs) | 59.9 ± 13.4 | 56.5 ± 13.0 | 59.0 ± 13.3 | 60.5 ± 13.2 | 63.6 ± 13.1 | <0.01 |
| Men | 45.2% | 59.5% | 41.7% | 40.5% | 38.9% | <0.01 |
| Black race | 9.2% | 14.1% | 9.3% | 7.0% | 6.2% | <0.01 |
| Body mass index (kg/m 2 ) | 27.6 ± 5.5 | 27.2 ± 5.3 | 27.4 ± 5.6 | 28.6 ± 5.4 | 28.1 ± 5.5 | <0.01 |
| Hypertension | 43.8% | 36.3% | 37.8% | 42.6% | 59.6% | <0.01 |
| Smoking | 23.0% | 24.4% | 23.2% | 22.4% | 22.0% | 0.61 |
| Hypercholesterolemia | 59.9% | 59.0% | 56.3% | 61.5% | 63.2% | 0.04 |
| Coronary artery disease | 9.8% | 6.7% | 7.6% | 11.6% | 13.5% | <0.01 |
| Myocardial infarction | 5.4% | 3.3% | 3.3% | 6.1% | 9.4% | <0.01 |
| Family history of heart attack | 11.1% | 8.9% | 10.4% | 11.1% | 14.1% | 0.008 |
| Heart failure | 2.8% | 1.0% | 2.1% | 3.0% | 5.4% | <0.01 |
| Stroke | 2.9% | 1.8% | 1.8% | 4.0% | 4.1% | <0.01 |
| Diabetes | 10.9% | 9.6% | 10.3% | 10.2% | 13.8% | 0.02 |
| High-density lipoprotein (mg/dl) | 51.0 ± 16.3 | 51.2 ± 16.0 | 51.6 ± 16.3 | 50.5 ± 16.0 | 50.5 ± 16.8 | 0.10 |
| Serum cholesterol (mg/dl) | 218.3 ± 44.1 | 214.2 ± 42.5 | 216.8 ± 44.0 | 219.2 ± 43.4 | 223.3 ± 45.9 | <0.01 |
| Serum triglycerides (mg/dl) | 161.9 ± 130.3 | 153.9 ± 126.3 | 157.5 ± 140.9 | 162.1 ± 117.7 | 174.8 ± 133.9 | <0.01 |
| Low-density lipoprotein (mg/dl) | 136.4 ± 38.6 | 134.1 ± 38.2 | 135.1 ± 38.3 | 137.0 ± 38.7 | 139.8 ± 38.9 | <0.01 |
| Glomerular filtration rate (ml/min) | 68.2 ± 15.6 | 71.4 ± 15.6 | 68.7 ± 15.1 | 67.2 ± 14.9 | 65.3 ± 16.2 | <0.01 |
| Serum calcium (mmol/L) | 1.23 ± 0.05 | 1.24 ± 0.05 | 1.23 ± 0.05 | 1.23 ± 0.05 | 1.23 ± 0.05 | 0.04 |
| Serum potassium (mEq/L) | 4.07 ± 0.35 | 4.10 ± 0.32 | 4.08 ± 0.35 | 4.07 ± 0.36 | 4.02 ± 0.38 | <0.01 |
| QRS duration | 96.0 ± 10.0 | 94.9 ± 10.2 | 95.6 ± 9.9 | 96.5 ± 9.9 | 97.2 ± 9.9 | <0.01 |
| Left ventricular hypertrophy | 7.5% | 2.4% | 5.3% | 8.1% | 14.8% | <0.01 |
Over a follow-up period of 13.5 ± 3.8 years, 1,226 CV deaths (15.5%) were reported ( Table 3 ). A multivariate model of ST-segment amplitude in lead aVR and traditional risk factors demonstrated ST-segment elevation (>8 μV) in lead aVR to be predictive of CV mortality ( Supplementary Table 1 ). Similarly, a multivariate model of T-wave amplitude in lead aVR and traditional risk factors demonstrated T-wave amplitude >−0.1 mV to be predictive of CV mortality ( Supplementary Table 2 ). Multivariate analysis of ST-segment and T-wave amplitude in lead aVR and traditional risk factors demonstrated T-wave amplitude in lead aVR >−0.1 mV to be independently predictive of CV mortality, whereas any change in ST-segment amplitude was not significantly associated with CV mortality. The reported hazard ratios for T-wave amplitude in lead aVR of −0.1 to 0 mV and >0 mV were 1.6 (95% confidence interval [CI] 1.2 to 2.2, p <0.01) and 3.3 (95% CI 2.1 to 5.3, p <0.01), respectively, considering T-wave amplitude <−0.2 mV as the referent ( Table 3 , Figure 1 ). Other traditional risk factors demonstrating associations with CV mortality on multivariate analysis included older age, male gender, hypertension, smoking, coronary artery disease, stroke, and diabetes.
| Variable | Hazard Ratio (95% CI) for CV Mortality (n = 1,226) | p Value |
|---|---|---|
| Older age | 1.12 (1.10–1.13) | <0.01 |
| Male gender | 1.38 (1.13–1.69) | <0.01 |
| History of hypertension | 1.48 (1.15–1.89) | <0.01 |
| Current smoking | 2.08 (1.67–2.60) | <0.01 |
| History of coronary artery disease | 1.79 (1.51–2.12) | <0.01 |
| History of stroke | 1.81 (1.42–2.31) | <0.01 |
| History of diabetes | 1.52 (1.26–1.84) | <0.01 |
| ST-segment amplitude in lead aVR (mV) | ||
| Quartile 1: −0.202 to −0.020 (n = 2,024) | Referent | |
| Quartile 2: −0.019 to −0.006 (n = 2,057) | 0.89 (0.62–1.28) | 0.53 |
| Quartile 3: −0.005 to 0.007 (n = 1,928) | 0.98 (0.75–1.28) | 0.89 |
| Quartile 4: 0.008 to 0.202 (n = 1,919) | 1.14 (0.83–1.55) | 0.42 |
| T-wave amplitude in lead aVR (mV) | ||
| <−0.2 (n = 3,048) | Referent | |
| −0.2 to −0.1 (n = 3,417) | 1.27 (0.99–1.64) | 0.06 |
| −0.1 to 0 (n = 1,288) | 1.66 (1.24–2.21) | <0.01 |
| >0 (n = 175) | 3.37 (2.11–5.36) | <0.01 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
