Electrocardiographic Predictors of Cardiovascular Outcome in Women

The National Heart, Lung, and Blood Institute-Sponsored Women’s Ischemia Syndrome Evaluation (WISE) Study

Study design

The WISE study, a National Heart, Lung, and Blood Institute-sponsored four-center study, aims to improve diagnostic testing in the evaluation of ischemic heart disease in women. Between 1996 and 2000, 954 women with suspected myocardial ischemia that prompted clinical referral for diagnostic coronary angiography were enrolled. Each site obtained institutional review board approval and participant consent before testing. Major exclusion criteria for the WISE study were comorbidity that would compromise one-year follow up, pregnancy, contraindications to provocative diagnostic testing, cardiomyopathy, New York Heart Association functional class III to IV CHF, recent myocardial infarction (MI), significant valvular or congenital heart disease, and a language barrier to questionnaire testing. Details of the design of the WISE study have been published elsewhere (5).

Baseline evaluation included a physical examination and collection of clinical and laboratory data, including use of medication. Age and self-reported risk factors (history of hypertension, diabetes, cigarette smoking, any family history of coronary artery disease [CAD], and history of CHF) were recorded at study entry. For analyses, dyslipidemia was defined as use of lipid-lowering medication (statins and others), a low-density lipoprotein cholesterol value of ≥130 mg/dl for women without any angiographic evidence of CAD, or a low-density lipoprotein cholesterol level of ≥100 mg/dl for those with angiographic evidence of CAD. Cholesterol levels were analyzed at the WISE lipid core laboratory at the Cedars Sinai Medical Center, which is enrolled in the Centers for Disease Control and Prevention lipid standardization program.

This report examined 143 women from the University of Pittsburgh WISE site who had complete angiographic and baseline data and who had ECGs that were stored digitally. Women with left or right bundle branch block patterns or with non-specific ventricular conduction abnormalities on their surface ECG were excluded from analyses.

ECGs

A resting 12-lead ECG was performed at study enrollment. The ECGs were masked to subject identification, digitized at acquisition, and sent to a core laboratory for interpretation using previously reported methods for ECG analyses (6,7). Measured variables included the heart rate and the PR, QRS, and QT intervals (both before and after adjustment for gender and race). We also measured ECG parameters that are infrequently analyzed, such as the root mean square of heart rate variations; the J point, ST-segment, and T-wave amplitudes in various leads; the QRS-T angle; the non-dipolar components of the QRS and T waves; the third component as well as the ratio of the third to second component of the Chebyshev waveform vector magnitude; the elevation and azimuth of mean T vector; the mean T vector spatial magnitude; and the interval from the first to second inflection point of the T-wave. A full description of the derivation methods for these parameters is described in detail elsewhere (6–8). Variables were selected for use in these analyses because of their association with repolarization, risk assessment, or myocardial ischemia. The spatial angle between the mean QRS and T vectors (QRS-T angle) was derived from waveform vector analysis, similar to principal component analysis or singular value decomposition (6). The amplitude of the ST-segment in precordial lead V₅60 ms after the end of the QRS complex, measured in microvolts (ST60V₅), has been previously associated with myocardial ischemia (8).

Angiographic data

All women in the WISE study underwent clinically indicated coronary angiography. Quantitative analysis was performed off-line at the WISE angiographic core laboratory at Rhode Island Hospital according to methods that have been previously published (9). Significant CAD was defined as the presence of at least one angiographic stenosis ≥50% of the luminal diameter. The CAD severity score was defined as an aggregate of percent luminal stenosis, extent and location of stenosis, and degree of collateral vessels (9).

Follow-up

An experienced nurse and/or physician collected in person or by telephone interview follow-up data at six weeks and then yearly thereafter. Each woman was queried for the occurrence of adverse outcomes such as hospitalization for angina, CHF, stroke, vascular conditions, MI, and death. When an adverse outcome was identified, the referring physician was contacted for confirmation, dates, and documentation.

Statistical analyses

Clinical and demographic data are summarized as means and standard deviations for clinical variables and as frequencies for categorical variables. Electrocardiographic variables are presented as medians with the interquartile range shown. Data also are presented as event rates for the median split for each of the ECG parameters. Comparisons of ECG variables in women with and without adverse events were done using two sample signed rank Wilcoxon tests. Coronary artery disease was defined as a dichotomous variable (CAD/no CAD) and as a continuous variable (severity score). Univariate and multivariable Cox proportional hazards regression models were used to identify predictors of adverse events. Cross correlations between ECG variables were examined before running any multivariate analyses. These events were defined as composite end points in two ways: 1) CHF, non-fatal MI, or death or 2) hospitalization for angina, CHF, non-fatal MI, or death. Variables were chosen for entry into the multivariable Cox models based on univariate associations using a stepwise procedure. For the Cox proportional hazards model, the hazard ratio (HR) and 95% confidence intervals (CIs) were calculated. Hazard ratios are shown per 10 units for ease of understanding. All tests were two-sided, and p values <0.05 were considered statistically significant. All statistics were analyzed using SAS version 8.2 (SAS Institute, Cary, North Carolina).

Study population

Table 1shows the baseline demographic data and CAD risk factor profiles of the 143 women included in our analysis. Primary events, defined as death, CHF, or non-fatal MI, were documented in 18 (13%) during a mean follow-up of 3.3 ± 1.6 years.

Parameters and events

Because we examined the event rates using the median split of the ECG variables, median, interquartile range, and cardiovascular event rates for women below and above the median are presented in Table 2.Only those variables with p values >0.1 are shown. In both groups, ECG parameters of heart rate and PR, QRS, and QTrr intervals were within previously reported normal limits for women. However, women with documented death, CHF, or non-fatal MI during follow-up had significantly higher (mean ± SD) heart rates (80 ± 13 beats/min vs. 71 ± 14 beats/min, p = 0.01) and wider QRS-T angles (78 ± 31° vs. 50 ± 26°, p = 0.0008) on their baseline ECG compared with women with no events. They also had a significantly more depressed ST-segment in the lateral precordial ECG leads V₅(−14 ± 61 μV vs. 11 ± 52 μV, p = 0.03) and V₆(−0.78 ± 90 μV vs. −19.8 ± 51 μV, p = 0.03). There was a trend toward wider QRS complexes (93 ± 12 ms vs. 87 ± 12 ms, p = 0.08) and QTrr intervals (437 ± 36 ms vs. 420 ± 21 ms, p = 0.05) in women with events.

Using an alternative composite end point of death, CHF, non-fatal MI, or angina requiring hospitalization (62 [43%] women), the ECG findings were similar and the QRS duration was significantly longer (mean ± SD) in the women with compared with those without events (91 ± 11 ms vs. 85 ± 13 ms, p = 0.005).

Univariate analysis

Univariate analysis of all ECG parameters are shown in Table 3.The correlation between significant univariate ECG predictors is shown in Table 4.Clinical and ECG predictors of outcome in the study population are presented in Table 5.Clinical characteristics that predicted the occurrence of death, CHF, or non-fatal MI included a history of diabetes mellitus (HR 3.71; 95% CI 1.44 to 9.59; p = 0.007), dyslipidemia (HR 9.21; 95% CI 1.21 to 70.20; p = 0.03), hypertension (HR 3.83; 95% CI 1.11 to 13.25; p = 0.03), and CAD measured by the angiographic severity score (HR 1.27 per 10-U increase in score, 95% CI 1.00 to 1.62; p = 0.05).

Univariate analyses also demonstrated that ECG predictors of events included wider QRS (HR = 1.60 per 10-ms widening, p = 0.03) and QTrr (HR = 1.25 per 10-ms widening, p = 0.003) intervals, higher heart rate (HR = 1.49 per 10 beats/min increase in heart rate, p = 0.01), and a wider QRS-T angle (HR = 1.36 per 10° increase in angle, p = 0.0002) (Table 5). Similar results were obtained when the composite event was defined as death, CHF, non-fatal MI, or angina leading to hospitalization (data not shown).

Multivariate analysis

A Cox regression model was used for multivariate analyses of clinical and ECG predictors of cardiovascular events (Table 5). Variables entered into this stepwise regression model included a diagnosis of diabetes mellitus, dyslipidemia, hypertension, and ECG variables of QRS-T angle, QRS duration, QTrr, and ST60V₅. These analyses demonstrated that independent predictors of death, CHF, and non-fatal MI included a wider QRS-T angle (HR = 1.50 per 10° increase in the angle, p = 0.0005), a wider QRS complex (HR = 1.75 per 10-ms widening, p = 0.004) intervals, a longer QTrr (HR = 1.46 per 10-ms increase in duration, p = 0.0004), a more depressed ST segment in precordial lead V₅(HR = 0.80 per 10 μV of depression in the ST-segment, p = 0.0002), and a higher angiographic severity score (HR = 1.40 per 10-U increase in score, p = 0.02).

Study limitations

It should be recognized that there are several potential limitations to our study. First, our study is a prospective analysis of the clinical and ECG predictors of cardiovascular events in women with suspected myocardial ischemia. It therefore focuses on a selected symptomatic group of patients, and the results presented here may not apply to other groups of women. Second, our patient population consisted solely of women and, therefore, our results do not necessarily apply to men. Finally, each patient in this study had one 12-lead surface ECG obtained at baseline. We recognize that some of the ECG parameters measured may not be constant or stable over time, which is a possibility that can potentially affect the predictive utility of these parameters. Nevertheless, our study demonstrates that several uncommonly used ECG parameters that are derived from a single random 12-lead ECG are predictive of subsequent cardiovascular events in women with suspected myocardial ischemia, independent of the presence and angiographic severity of preexisting angiographic CAD.

Conclusions

The present study supports the hypothesis that several ECG parameters are predictive of subsequent cardiovascular events in a relatively healthy population of women with suspected myocardial ischemia. A wider QRS-T angle, a wider QRS duration, and a longer QTrr were all found to be statistically significant independent predictors of death, CHF, and non-fatal MI. The novelty of this study lies in the fact that abnormalities of these ECG parameters were predictive of events in a relatively healthy population of women even when their values were within the accepted normal range. The results of this study raise the issue of incorporation of these sophisticated ECG parameters as part of a global risk factor score. Future prospective studies that evaluate a general asymptomatic population are needed to extend the findings of our study. If verified, these measurements could provide a noninvasive and relatively inexpensive means of assessing cardiovascular risk.