Clinically Silent Electrocardiographic Abnormalities and Risk of Primary Cardiac Arrest Among Hypertensive Patients
Background Whether continuous ECG indexes that reflect the severity of left ventricular hypertrophy (LVHI), myocardial injury (CIIS), and QT-interval prolongation (QTI) are associated with the risk of primary cardiac arrest among hypertensive patients, independent of conventional binary ECG criteria, remains unknown.
Methods and Results We conducted a population-based case-control study among patients who were free of clinically recognized heart disease and who received care at a health maintenance organization. Cases (n=131) were treated hypertensive patients who had had a primary cardiac arrest between 1977 and 1990. Controls (n=562) were a stratified random sample of treated hypertensive patients. Resting ECGs were reviewed to estimate the severity of left ventricular hypertrophy, myocardial injury, and QT-interval prolongation on the basis of the algorithms of the Novacode ECG classification system. After adjustment for other risk factors and binary ECG criteria for the abnormalities, the LVHI, CIIS, and QTI scores were directly related to the risk of primary cardiac arrest. In a comparison of the 80th with the 20th percentile score for the LVHI, the risk was increased 40% (odds ratio, 1.4; 95% CI, 1.0 to 2.0); for the CIIS, the risk was increased 70% (odds ratio, 1.7; 95% CI, 1.2 to 2.5); and for the QTI, the risk was increased 80% (odds ratio, 1.8; 95% CI, 1.3 to 2.7).
Conclusions Our findings suggest that continuous ECG indexes that reflect left ventricular hypertrophy, myocardial injury, and QT-interval prolongation are directly related to the risk of primary cardiac arrest among hypertensive patients without clinically recognized heart disease. Binary ECG criteria may underestimate the prognostic importance of these pathophysiological abnormalities.
On the basis of conventional criteria, ECG findings of left ventricular hypertrophy, myocardial injury, and QT-interval prolongation are associated with an increased risk of sudden cardiac death.1 2 3 These pathophysiological abnormalities may account, in part, for the increased risk of sudden cardiac death among hypertensive patients.4 5 However, these binary criteria detect only severe abnormalities of cardiac structure and electrophysiological function; moderate abnormalities also may increase the risk of life-threatening arrhythmias among hypertensive patients.
ECG indexes have been developed to assess the severity of left ventricular hypertrophy, myocardial injury, and QT-interval prolongation on a continuous scale.6 7 8 Because the relation of these pathophysiological abnormalities with the risk of primary cardiac arrest may be graded, continuous ECG indexes that reflect the severity of these abnormalities may be directly related to the occurrence of primary cardiac arrest. A recent case-control study provided the opportunity to determine whether clinically silent resting ECG findings that reflect the severity of left ventricular hypertrophy, myocardial injury, and QT-interval prolongation, assessed with the use of continuous ECG indexes, were related to the risk of primary cardiac arrest, independent of conventional binary criteria, among treated hypertensive patients.
Selection of Case and Control Subjects
The design of this study has been described in detail elsewhere.9 Briefly, we conducted a population-based case-control study among treated hypertensive patients who received medical care at GHC, a large health maintenance organization. With information from the emergency medical services systems of Seattle and suburban King County (Washington), the Washington State death files, and the GHC enrollment files, we identified all incident cases of out-of-hospital primary cardiac arrest among GHC enrollees during the period of 1977 through 1990. Primary cardiac arrest was defined operationally as a sudden pulseless condition in the absence of a known noncardiac condition that could account for cardiac arrest.10 We used the GHC computerized pharmacy files to identify the cases who received antihypertensive drugs during the year before the event. For each case, the index date was the date of primary cardiac arrest.
Control subjects were treated hypertensive patients who were under care at GHC during the study period. They were randomly selected from the GHC pharmacy file stratified by age (decade), sex, and calendar year of treatment. Each control subject was assigned an index date that was chosen at random from the distribution of index dates among the cases.
We excluded case and control subjects who were <30 years or >79 years of age or who were not residents of King County. We also excluded subjects who had had clinically recognized heart disease, such as physician-diagnosed myocardial infarction, angina pectoris, congestive heart failure, arrhythmias, and valvular heart disease, or other life-threatening conditions, such as end-stage lung, liver, or renal disease or cancer, before their index dates, on the basis of an ambulatory medical record review. We also confirmed through the medical record review that all subjects were under treatment for hypertension on their index dates.
Measurement of Exposures
Ambulatory-care medical records were reviewed for all cases and control subjects. Copies of the most recent ECG before the index date were obtained and sent to the EPICARE Center (Bowman Gray School of Medicine) for reading. ECGs were available for 131 of 173 cases (76%) and 562 of 768 control subjects (73%). Analyses of subjects with and without ECGs suggested that subjects who had had an ECG were older (mean difference, 3.9 years for cases and 2.3 years for controls), had a longer duration of hypertension (mean difference, 2.7 years for cases and 2.5 years for controls), and were more likely to be diabetic (for cases only, a difference in prevalence of 15%) than were subjects who had not had an ECG (P<.05 for each comparison). However, other clinical characteristics were not significantly associated with the availability of an ECG in the medical record, including sex, race, smoking, pretreatment systolic and diastolic blood pressures, posttreatment systolic and diastolic blood pressures, antihypertensive drug therapy, and serum glucose, potassium, and creatinine levels. The time interval between the ECG and the index date was similar for cases and control subjects; for case subjects, the mean (SD) interval was 4.4 years (4.7 years), and for control subjects, the mean interval was 4.6 years (4.2 years) (for difference between the mean values, P=.7).
Blinded to case-control status, research associates used standard criteria to code abnormalities. ECGs were coded according to the MC.11 Using the methods of the Multiple Risk Factor Intervention Trial, we combined specific MCs to estimate the prevalence of discrete ECG abnormalities5 : left ventricular hypertrophy was defined by the presence of high-amplitude R waves with ST-T–wave changes (MC 3.1, 3.3, and 4.1-4.3 or 5.1-5.3), myocardial injury by the presence of ST-T–wave abnormalities without high-amplitude R waves and Q and QS waves (MC 4.1-4.3 or 5.1-5.3 and 1.1-1.3), and conduction abnormalities by the presence of bundle-branch blocks (MC 7.1, 7.2, and 7.4).
In addition, we used algorithms derived from the Novacode ECG classification system6 12 13 14 to estimate the severity of left ventricular hypertrophy (adjusted for weight), prior myocardial infarction/injury, and QT-interval prolongation (adjusted for heart rate). Information from multiple ECG features was combined to calculate scores for three continuous ECG indexes: the LVHI, CIIS, and QTI.6 7 8
Because ECGs were coded visually, we simplified the Novacode algorithm used to estimate the LVHI (adjusted for weight) in the following manner: LVHI=k exp[0.01×(R-wave aVL+S-wave V3+R-wave V5)+3.6×QRS duration−0.03×T-wave V6], where k=56.6 for men and 60 for women; R-, S-, and T-wave amplitudes are given in standardized millimeters at 10 mm/mV, and QRS duration is given in seconds. This formula yields an LVHI score of 100 for subjects with average ECG amplitudes and without echocardiographic LVH. We adjusted the measured QT interval for heart rate using the QTI formula derived from a large population-based sample: QTI=[QT×(heart rate+100)/656].8 A QTI score of >110 indicates that the QT interval exceeds by >10% the predicted QT interval for a given heart rate. Twelve ECG features were used to compute the CIIS: five T-wave amplitude measurements; four Q-wave duration measurements or Q:R-amplitude ratios; and three R- or S-amplitude measurements.6 13 We used a previously described visual ECG reading coding form to score the CIIS.15
From the ambulatory care medical record review, information was collected on age, sex, race, number of visits during the year before the index date, duration of hypertension, pretreatment and posttreatment systolic and diastolic blood pressures, current smoking, diabetes, weight, and most recent serum glucose, potassium, and creatinine levels. We used the GHC computerized pharmacy database to assess the type (thiazide diuretics, thiazide diuretics with potassium-sparing agents, β-adrenergic–blocking agents, other antihypertensive drugs) and dose (for thiazide diuretics) of antihypertensive drug therapy used at the index date.
We used unconditional logistic regression to estimate the risk associated with a higher score for each ECG index after adjustment for the other clinical characteristics. Preliminary analyses suggested that the correlations among the ECG indexes were small and that there was little evidence of a nonlinear relation between each of the ECG indexes and the risk of primary cardiac arrest. For these reasons, the logistic models included the linear terms for each ECG index, considered individually and simultaneously. We also determined whether the addition of interaction terms for each ECG index and sex, current smoking, and antihypertensive therapy improved the fit of the model. The statistical significance associated with the addition of a variable to the model was based on the likelihood ratio statistic. Fifteen cases and 67 control subjects with missing data for an ECG feature or a covariate were excluded from the multivariate analyses.
The cases and controls were similar demographically (Table 1⇓): the mean age of cases and controls was 66 and 64 years, respectively; 56% of cases and 54% of controls were male; and 94% of cases and 96% of controls were white. The prevalences of current smoking, diabetes, and treatment with moderate or high-dose thiazide diuretic therapy (50 or 100 mg/d) were higher among case subjects than among control subjects, and the pretreatment systolic blood pressure, resting heart rate, and most recent serum glucose and creatinine levels were higher among cases than among control subjects. The mean number of visits during the prior year, duration of hypertension, weight, serum potassium level, and prevalences of treatment with β-blocker therapy and nondiuretic/non–β-blocker therapy alone were similar among cases and controls.
ECG evidence of left ventricular hypertrophy, myocardial injury, and conduction abnormalities based on the MCs was more common among cases than among controls (Table 2⇓). The LVHI, CIIS, and QTI were also greater among cases than control subjects (Table 2⇓). With increasing LVHI, CIIS, and QTI scores, there was an increase in the risk of primary cardiac arrest: compared with the lowest tertile for each index, the middle tertile was associated with a modest increase in risk and the upper tertile was associated with a larger increase in risk (Table 3⇓). However, for the LVHI, CIIS, and QTI, only 35%, 48%, and 47% of the cases, respectively, were in the highest tertile of the index.
After adjustment for the other clinical characteristics, including age, sex, current smoking, diabetes, treated diastolic blood pressure, resting heart rate (for LVHI and CIIS), and antihypertensive therapy, each of the three ECG indexes (LVHI, CIIS, and QTI) was directly related to the risk of primary cardiac arrest (Table 4⇓). In a comparison of the 80th with the 20th percentile scores on the basis of the logistic models, the risk was increased 40% for the LVHI, 70% for the CIIS, and 80% for the QTI.
Further adjustments for weight, serum glucose, creatinine, and potassium levels altered the results only slightly (data not shown). There was little evidence that sex, current smoking, or antihypertensive therapy modified the relation between each ECG index and the risk of primary cardiac arrest (data not shown). There also was little evidence to suggest interactions among the ECG indexes (data not shown).
The ECG indexes were related to the risk of primary cardiac arrest even after further adjustment for binary ECG criteria for these abnormalities on the basis of the MCs (Table 4⇑). Adjustment for high-amplitude R waves with ST-T–wave changes altered the relation between the LVHI and the risk of primary cardiac arrest only slightly; adjustment for ST-T–wave changes both with and without major Q and QS waves had little effect on the risk associated with a higher CIIS, and adjustment for bundle-branch blocks had trivial effects on the association between the QTI and the risk of primary cardiac arrest.
When the three ECG indexes were considered simultaneously in the logistic regression model adjusted for other clinical characteristics, both the 80th percentile of the CIIS and the QTI were associated with a twofold increase in the risk of primary cardiac arrest compared with the 20th percentile (Table 4⇑). In contrast, after adjustment for the other clinical characteristics and ECG indexes, the LVHI was associated with only a slight increase in the risk of primary cardiac arrest (Table 4⇑).
Our findings suggest that among treated hypertensive patients without clinically recognized heart disease, ECG indexes that reflect the severity of left ventricular hypertrophy, myocardial injury, and QT-interval prolongation were directly related to the risk of primary cardiac arrest. Even ECG findings consistent with moderate abnormalities of cardiac structure and function were associated with an increased risk of primary cardiac arrest. The relations were altered only slightly by adjustment for other clinical characteristics. Of note, both the CIIS and the QTI were related to the risk of primary cardiac arrest after adjustment for the other clinical characteristics and ECG indexes. Finally, the ECG indexes were related to the risk of primary cardiac arrest even after adjustment for the binary ECG criteria for these abnormalities.
The design of our study precludes a full evaluation of the clinical usefulness of the ECG indexes. Because cardiac arrest is a rare event among hypertensive patients without prior clinically recognized heart disease, we used a population-based case-control study design to evaluate potential predictors of cardiac arrest. Although our study was retrospective, we identified all cases of cardiac arrest among a cohort of ≈20 000 hypertensive patients who received medical care at a large health maintenance organization facility during a 14-year period; both the information used to assess clinical characteristics and ECGs were recorded before an index date; for cases, the date of the occurrence of cardiac arrest was used. Because the study was population based, the selection of control subjects (noncases) was straightforward: controls were randomly sampled, pharmacologically treated hypertensive patients who were demographically similar to cases and under care during the same period of time. However, because we collected information on only a subset of the hypertensive patients who did not experience cardiac arrest, we are not able to estimate directly the predictive value (positive or negative) and receiver-operator characteristic curve of the ECG indexes.
We used copies of ECGs that were available in the ambulatory care medical record, and we used visual coding to interpret the ECGs. Information on the specific indication for the ECG was not collected. Although we excluded subjects who had had a physician diagnosis of heart disease before their index date, it is possible that in some cases ECGs were obtained as part of the evaluation of suspected coronary heart disease or arrhythmias. On the other hand, the ECG was commonly obtained as part of the initial evaluation of a hypertensive patient. Although subjects with ECGs available were older, had a longer duration of hypertension, and were more likely to be diabetic (for cases only) than were subjects without an ECG, the magnitude of the differences was small, and these clinical characteristics were taken into account in our multivariate analyses. Nevertheless, assessment of the indication for the ECG is needed to evaluate more fully the clinical usefulness of these ECG indexes.
Our findings extend the results of prior studies. In the Framingham Heart Study, conventional ECG criteria for left ventricular hypertrophy and intraventricular conduction delays were related to the risk of sudden cardiac death, but the prevalences of these abnormalities were low.3 Similarly, ECG evidence of LVH, myocardial injury, and conduction delays, based on computer coding of ECGs with the use of the MCs, was related to coronary mortality among the hypertensive men who participated in the Multiple Risk Factor Intervention Trial.5 In the NHANES Epidemiologic Follow-up Study, ECG estimates of left ventricular mass and myocardial injury, based on continuous ECG indexes, were related to the risk of coronary heart disease mortality in the population as a whole.16
The findings that we report suggest that ECG indexes that reflect the severity of left ventricular hypertrophy, myocardial injury, and QT-interval prolongation were directly related to the risk of primary cardiac arrest among hypertensive patients without prior clinically diagnosed heart disease; there was a modest increase in risk associated with a higher score for each index that was independent of other clinical characteristics and conventional ECG criteria. For this reason, analyses based on conventional binary ECG criteria may underestimate the importance of these pathophysiological abnormalities as potential determinants of primary cardiac arrest. Taken together with prior pathological and electrophysiological studies, the findings suggest that not only the presence but also the severity of abnormalities of cardiac structure and electrophysiological function account, in part, for the increased risk of primary cardiac arrest observed among treated hypertensive patients.
Prospective cohort studies are needed to verify these findings and to more fully explore the clinical usefulness of these indexes in the prediction of primary cardiac arrest among hypertensive patients. Although the results suggest that ECG indexes may be better predictors than the traditional binary ECG criteria that are used to assess left ventricular hypertrophy, myocardial damage, and conduction abnormalities, clinicians should be aware that both the sensitivity of these indexes and the increase in absolute risk associated with the ECG abnormalities reflected by the available indexes remain modest.
Selected Abbreviations and Acronyms
|CIIS||=||Cardiac Infarction/Injury Score|
|GHC||=||Group Health Cooperative of Puget Sound|
|LVHI||=||Left Ventricular Hypertrophy Index|
This work was supported by grant HL-42456-03 from the National Heart, Lung, and Blood Institute.
Reprint requests to David S. Siscovick, MD, MPH, Cardiovascular Health Research Unit, Metropolitan Park 2 Building, Suite 1360, 1730 Minor Ave, Seattle, WA 98101.
- Received November 29, 1995.
- Revision received March 21, 1996.
- Accepted March 26, 1996.
- Copyright © 1996 by American Heart Association
Kannel WB, Gordon T, Castelli WP, Margolis JR. Electrocardiographic left ventricular hypertrophy and risk of coronary heart disease: the Framingham Study. Ann Intern Med. 1970;72:813-822.
Kannel WB, Gordon T, Offutt D. Left ventricular hypertrophy by electrocardiogram: prevalence, incidence, and mortality in the Framingham Study. Ann Intern Med. 1969;71:89-101.
Rautaharju PM, LaCroix AZ, Savage DD, Haynes SG, Madans JH, Wolf HK, Hadden W, Keller J, Cornoni-Huntley J. Electrocardiographic estimate of left ventricular mass versus radiographic cardiac size and the risk of cardiovascular disease mortality in the Epidemiologic Follow-Up Study of the First National Health and Nutrition Examination Survey. Am J Cardiol. 1988;62:59-66.
Rautaharju PM, Warren JW, Jain U, Wolf HK, Nielsen CL. Cardiac infarction injury score: an electrocardiographic coding scheme for ischemic heart disease. Circulation. 1981;64:249-256.
Rautaharju PM, Warren JW, Calhoun HP. Estimation of QT prolongation: a persistent avoidable error in computer electrocardiography. J Electrocardiol. 1990;23:111-117.
Blackburn H, Keys A, Simonson E, Rautaharju P, Punsar S. The electrocardiogram in population studies: a classification system. Circulation. 1969;21:1160-1175.
Rautaharju PM, MacInnis PJ, Warren JW, Wolf HK, Rykers PM, Calhoun HP. Methodology of ECG interpretation in the Dalhousie Program NOVACODE ECG classification procedures for clinical trials and population health surveys. Methods Info Med. 1990;299:362-374.
Rautaharju PM, Calhoun HP, Chaitman BR. Novacode serial ECG classification system for clinical trials and epidemiological studies. J Electrocardiol. 1992;24:163-172.
Rataharju PM. Electrocardiogram in epidemiology and clinical trials. In: Macfarlane PW, Lawrie TDV, eds. Comprehensive Electrocardiography. New York, NY: Pergamon Books Ltd; 1988:1244.
Rautaharju PM, Zhou SH, Calhoun HP, Semples C, Madans JH, Cox SC, Hadden W. ECG manifestations of subclinical disease: prognostic importance. In: Macfarlane P, Rautaharju PM, eds. Electrocardiology '93. London, UK: World Scientific; 1994:304-308.