Competing cardiovascular outcomes associated with electrocardiographic left ventricular hypertrophy: the Atherosclerosis Risk in Communities study

Abstract

OBJECTIVE

Individuals with electrocardiographically-determined left ventricular hypertrophy (ECG LVH) are at risk for multiple cardiovascular disease (CVD) outcomes simultaneously. We sought to characterize the competing incidences for subtypes of first CVD events or non-CVD death in those with and without ECG LVH.

DESIGN

We included participants in the Atherosclerosis Risk in Communities (ARIC) study. ECG LVH was defined according to Sokolow-Lyon criteria. We used competing Cox models to compare hazards for diverse outcomes within groups (e.g., among those with ECG LVH) and for a given event between groups (ECG LVH versus no ECG LVH).

RESULTS

After 15 years, men with ECG LVH at baseline (N = 383) had cumulative incidence of first CVD events and non-CVD deaths of 29.2% and 6.1%, respectively (hazard ratio 4.86; 95% CI, 3.04–7.77). In men without ECG LVH (N = 6576) the incidence of any first CVD event and non-CVD death was 18.9% and 6.9%, respectively (hazard ratio 2.67; 2.39–2.98). Similar associations were observed in women (N = 381 with and N = 8187 without ECG LVH). Coronary heart disease (CHD) was the most common first event in men with ECG LVH (15.0%) and heart failure (HF) was the most common first event in women with ECG LVH (10.5%). After adjustment for risk factors including systolic blood pressure, any CVD event remained the most likely first event.

CONCLUSIONS

Among middle-aged individuals with ECG LVH, the most likely first events are CHD in men and HF in women; these results may have implications for preventive approaches.

BACKGROUND

Left ventricular hypertrophy (LVH) is most commonly a manifestation of end-organ damage due to hypertension.[1–3] When detected by the 12-lead electrocardiogram (ECG), LVH is strongly predictive of cardiovascular disease (CVD) including myocardial infarction (MI), sudden cardiac death, stroke, and congestive heart failure.[1,4,5] Other investigations have revealed that echocardiographically-determined LVH and increased left ventricular mass are associated with increased risk of cardiovascular disease.[6]

Individuals are at risk of death from competing non-cardiovascular events, particularly as they are followed for longer periods of time, and which are not accounted for in standard survival analyses. Furthermore, individuals at risk for CVD are at risk for multiple potential manifestations (e.g., coronary heart disease, stroke, heart failure) simultaneously, not independently. This distinction is important as different CVD events are associated with greater or lesser morbidity and mortality and they may confer differential risks for subsequent CVD events. Further, different strategies may be needed to prevent diverse first CVD events. No prior study has examined which CVD event occurs first after ECG LVH diagnosis.

The method described by Lunn and McNeil utilizes competing Cox models to compare hazard ratios and confidence intervals for diverse competing outcomes within a group (e.g., those with ECG LVH).[7] Fine and Gray described a complementary method that facilitates comparison of hazard ratios and confidence intervals for a given event,[8] such as the risk of specific CVD events, between different strata (e.g., comparing those with and without ECG LVH). The objective of our study is to determine the competing risks for different first CVD events and non-CVD deaths in individuals with ECG LVH.

METHODS

Study population

We included subjects from the Atherosclerosis Risk in Communities (ARIC) study and details of the study design have been published.[9] We used a limited access dataset from the National Heart, Lung, and Blood Institute (NHLBI). The study population comprises 15,792 individuals recruited in 1987–1989 from Forsyth County, North Carolina, Jackson, Mississippi, Minneapolis, Minnesota, and Washington County, Maryland. The overall cohort was aged 45–64 years at enrollment and consisted of 55% women and 27% African-American. All subjects were free of clinical cardiovascular disease at the baseline examination.

Electrocardiography

Patients underwent standard supine 12-lead electrocardiograms, with each tracing consisting of ten seconds of each of the 12 leads simultaneously. Electrocardiographic data processing, monitoring, and quality control have been described elsewhere.[10] All electrocardiograms were classified by the Minnesota code at a single reading center.[11] Electrocardiographic left ventricular hypertrophy (ECG LVH) was defined according to the Sokolow-Lyon criteria: sum of S wave in V1 and R wave in V5 or V6 ≥ 3.5 mV (35 mm) and/or R wave in aVL ≥ 1.1 mV (11 mm).[12] For this analysis, ECG LVH status was determined at the initial baseline examination. Continuous ECG measurements including QRS interval were also recorded.

Baseline measurements

Race and smoking status were determined by self-report. Resting systolic and diastolic blood pressure was measured by a random-zero sphygmomanometer. Diabetes was defined as fasting glucose ≥ 126 mg/dl, random glucose ≥ 200 mg/dl, previous diagnosis of diabetes, or pharmacologic therapy for diabetes. Body mass index (BMI) was calculated as the weight in kilograms divided by the height in meters squared.

Ascertainment of outcomes

Events were ascertained as previously described.[9] The duration of follow up was 15 years. Deaths were investigated by a review of death certificates, coroner records, and contact with physicians and patient families (when available). Outcomes of interest were death due to coronary heart disease, nonfatal myocardial infarction, stroke or transient ischemic attack, heart failure, other cardiovascular death, and non-cardiovascular death.

Statistical analysis

Baseline characteristics were compared by ECG left ventricular hypertrophy status using general linear models for continuous variables and chi-square tests for categorical variables. Sex differences in the prevalence of cardiovascular risk factors and in cardiovascular disease rates have been well described;[13–14] thus, we performed all analyses after stratification of the study sample by sex. All participants were followed until the occurrence of a CVD event, non-CVD death, or censoring for end of follow up at 15 years. Adjusted risks for CVD events were calculated separately for men and women stratified by ECG LVH status. Standard errors and 95% confidence intervals were estimated. We then determined the first event occurring during the follow up time, whether it was a CVD event or non-CVD death. If a CVD event occurred on the same day as the day of death, then the CVD event was coded as occurring first. When multiple CVD events were diagnosed as occurring on the same date, we arbitrarily assigned one as occurring first; for example, an individual was assigned with MI as occurring before heart failure (HF) if both were diagnosed on the same date. Competing Cox regression models were used to analyze the competing risks separately for men and women and by ECG LVH status. We used the data augmentation method as described by Lunn and McNeil[7] to fit Cox proportional hazards[15] models for all CVD events combined compared with non-CVD death as a first event. Standard Kaplan-Meier survival analyses are typically used in situations where one event is considered, and the time to event is considered the failure time. In the competing risks model we used in this study, a participant may fail from only one of the competing risks, and the time to the first event is considered the failure time. The hazards and the event-free survival are obtained from the augmented model and the competing cumulative incidence rate is the product of these two quantities. Therefore, we were able to estimate hazards ratios and cumulative incidences for competing CVD events compared with non-CVD death within a given group (e.g., those with ECG LVH).

In separate analyses, we used the method described by Fine and Gray[8,16] to estimate the subdistribution hazard separately for men and women and by ECG LVH status for each of five competing outcomes: 1) coronary heart disease death (CHD) or nonfatal myocardial infarction (NFMI); 2) heart failure (HF); 3) fatal or nonfatal stroke or transient ischemic attack (TIA); 4) other cardiovascular death; 5) and non-cardiovascular death. The Fine and Gray model is a modified Cox proportional hazards model that accounts for competing risks for different outcomes. The subdistribution hazards are modeled by keeping the competing risk observations in the risk set with diminishing weights. Thus, the effect estimated from the Fine and Gray model shows the current and real differences between the two groups in terms of subdistribution hazards. This model also utilizes time-dependent covariates to model the nonproportionality of hazards. We further adjusted for age, race, systolic blood pressure, QRS duration, total cholesterol, diabetes, and smoking status. R version 2.10.1 and its competing risk library were used for these analyses.

RESULTS

Baseline characteristics

The ARIC sample for this study included 6959 men and 8568 women, of whom 383 men and 381 women had ECG LVH. The median duration of follow up was 11.2 years and very few participants were lost to follow up with regards to vital status (< 1%). Baseline characteristics by sex and ECG LVH status are shown in Table 1. Participants with ECG LVH were older and more likely to be black. Higher systolic blood pressure (SBP) and higher BMI were associated with the presence of ECG LVH. Total cholesterol was associated with ECG LVH in men but not in women, and high-density lipoprotein (HDL) cholesterol was associated with ECG LVH in women but not in men. The prevalence of smoking tended to be higher in those without ECG LVH. Men and women with ECG LVH were more likely to have diabetes mellitus. The QRS interval was longer in both men and women with ECG LVH compared to men and women without ECG LVH. In addition, the QRS interval was greater in men with ECG LVH compared to women with ECG LVH (P < 0.001).

Table 1.

Characteristics of ARIC participants by sex and ECG LVH status

	Men			Women
	LVH+ (N = 383)	No LVH (N = 6576)	P value	LVH+ (N = 381)	No LVH (N = 8187)	P value
Age (years)	55.1 ± 5.7	54.6 ± 5.8	0.07	55.5 ± 5.8	53.7 ± 5.7	< 0.0001
African-American, n (%)	175 (46%)	1426 (22%)	< 0.0001	270 (71%)	2312 (28%)	< 0.0001
Systolic blood pressure (mm Hg)	133.6 ± 21.8	121.8 ± 17.5	< 0.0001	137.5 ± 24.9	119.6 ± 18.8	< 0.0001
Diastolic BP (mm Hg)	82.2 ± 13.2	74.9 ± 11.0	< 0.0001	81.0 ± 13.2	71.7 ± 10.8	< 0.0001
Body mass index (kg/m2)	29.6 ± 4.2	27.4 ± 4.2	< 0.0001	32.4 ± 6.2	27.7 ± 6.0	< 0.0001
Total cholesterol (mmol/L)	5.63 ± 1.10	5.45 ± 1.03	0.0009	5.70 ± 1.22	5.64 ± 1.12	0.32
HDL cholesterol (mmol/L)	1.13 ± 0.33	1.15 ± 0.36	0.23	1.36 ± 0.39	1.49 ± 0.45	< 0.0001
Current smoker, n (%)	59 (15%)	1871 (28%)	< 0.0001	56 (15%)	2078 (25%)	< 0.0001
Diabetes, n (%)	53 (14%)	616 (9%)	0.003	84 (22%)	787 (10%)	< 0.0001
Treated for Hypertension, n (%)	158 (42%)	1462 (22%)	< 0.0001	211 (55%)	2095 (26%)	< 0.0001
QRS duration (msec)	106 ± 18	101 ± 13	< 0.0001	100 ± 15	94 ± 11	< 0.0001

Values expressed as mean ± SD or N (row %)

ECG LVH = electrocardiographic left ventricular hypertrophy, HDL = high-density lipoprotein

Competing Cox regression models

Table 2 shows data from Cox models representing competing cumulative incidences of CVD events vs. non-CVD deaths, stratified by sex and ECG LVH status and using the methods of Lunn and McNeil on the left and Fine and Gray on the right. In men without ECG LVH, the cumulative incidence of any CVD as a first event was 18.9% and the incidence of non-CVD death was 6.9%, for a hazards ratio of 2.67 (95% CI, 2.39–2.98). In men with ECG LVH, the cumulative incidence of CVD events and non-CVD deaths was 29.2% and 6.1%, respectively, for a hazards ratio of 4.86 (95% CI, 3.04–7.77). Thus, any CVD event was more likely to occur first regardless of ECG LVH status, although the hazards ratio was greater in men with ECG LVH. The competing cumulative incidence curves for CVD events and non-CVD deaths are shown in Figure 1.

Table 2.

Summary table of Hazards Ratios and Competing Cumulative Incidences for CVD Events by LVH status

		Hazards Ratios and Competing Cumulative Event Incidences within group (Lunn and McNeil Method)				Hazards Ratios Between LVH-Sex groups (Fine and Gray Method)
		Men		Women		Men	Women
		LVH (−) N = 6576	LVH (+) N = 383	LVH (−) N=8187	LVH (+) N = 381	Hazards Ratio for Selected Event (LVH (+) vs. (−); LVH (−) as Referent)	Hazards Ratio for Selected Event (LVH (+) vs. (−); LVH (−) as Referent)
Model 1: unadjusted
Hazards Ratio for CVD vs Non-CVD Death Within Group		2.67 (2.39, 2.98)	4.86 (3.04, 7.77)	2.51 (2.20, 2.86)	8.36 (4.48, 15.6)
		Cumulative incidence, %
Non-CVD Death		6.9	6.1	4.1	3.2	0.84 (0.54, 1.30)	0.76 (0.41, 1.38)
All CVD events		18.9	29.2	10.7	26.3	1.66 (1.35, 2.04)	2.78 (2.23, 3.45)
	CHD/NFMI	10.7	15.0	4.5	10.2	1.41 (1.07, 1.89)	2.49 (1.77, 3.49)
	HF	5.3	9.7	4.2	10.5	1.96 (1.36, 2.83)	2.75 (1.94, 3.91)
	Stroke	2.7	4.2	1.8	5.3	1.70 (1.01, 2.84)	2.77 (1.70, 4.52)
	Other CVD death	0.3	0.3	0.2	0.3	0.85 (0.11, 6.36)	1.43 (0.19, 10.8)
Model 2: adjusted for age and race
All CVD event						1.54 (1.25, 1.90)	1.80 (1.44, 2.26)
Model 3: adjusted for age, race, and SBP
All CVD events						1.36 (1.10, 1.68)	1.43 (1.13, 1.81)
Model 4: adjusted for age, race, SBP, and QRS duration
All CVD events						1.25 (1.01, 1.55)	1.32 (1.04, 1.67)
Model 5: adjusted for age, race, SBP, diabetes, total cholesterol, smoking, and QRS duration
All CVD events						1.39 (1.12, 1.73)	1.37 (1.06, 1.76)

CVD = cardiovascular disease, LVH = left ventricular hypertrophy, CHD = coronary heart disease death, NFMI = nonfatal myocardial infarction, HF = heart failure, SBP = systolic blood pressure

Figure 1.

Cumulative incidence of events (occurring as first events) in men in ARIC by ECG LVH status

The relative hazards ratios for specific CVD events or non-CVD deaths for those with ECG LVH compared to those without ECG LVH are shown on the right side of Table 2. The most likely first event in men with and without ECG LVH was coronary heart disease death (CHD)/nonfatal myocardial infarction (NFMI), with an incidence of 15.0% and 10.7%, respectively. Thus, the relative hazards ratio for CHD/NFMI occurring first in men with ECG LVH compared to men without was 1.41 (95%, CI 1.07–1.89). The cumulative incidence of heart failure as a first event in men with and without ECG LVH was 9.7% and 5.3%, respectively; the relative hazards ratio was 1.96 (95% CI 1.36–2.83). Men with ECG LVH also appeared more likely to be diagnosed with stroke as a first event.

In women, the relative hazards ratios for any CVD event vs. non-CVD death occurring first was significantly greater in individuals with ECG LVH, compared to those without ECG LVH, as shown in Table 2 and Figure 2. Similar to men, the incidence of any CVD as a first event was higher in women with ECG LVH, and there was no statistically significant difference in non-CVD death as a first event between women with and without ECG LVH. Women with ECG LVH were more likely to experience any CVD event during the follow up period and this association remained significant in the analysis of specific events, including for CHD/NFMI, heart failure, and stroke. The most likely first event in women with ECG LVH was heart failure (10.5%), followed closely by CHD/NFMI (10.2%); the most likely first CVD event in women without ECG LVH was CHD/NFMI (4.5%).

Figure 2.

Cumulative incidence of events (occurring as first events) in women in ARIC by ECG LVH status

We adjusted for the effects of age and race in model 2 of Table 2. The hazards ratio for any CVD event occurring first in men with ECG LVH compared to those without was 1.54 (95% CI, 1.25–1.90); prior to adjustment for age and race the hazards ratio was 1.66 (95% CI, 1.35–2.04). In women, the adjusted hazards ratio for any CVD event occurring first in those with ECG LVH was 1.80 (1.44–2.26); prior to adjustment for age and race, the hazards ratio was 2.78 (95% CI, 2.23–3.45). Thus, a greater magnitude of attenuation was observed in women compared to men. Model 3 of Table 2 shows hazards ratios for any CVD event occurring first in individuals with ECG LVH compared to those without, adjusted for age, race, and SBP. The hazards ratios in men and women were further attenuated to 1.36 (95% CI, 1.10–1.68) and 1.43 (95% CI, 1.17–1.94), respectively. In Model 4 of Table 2, we adjusted for age, race, SBP, and QRS duration, and the hazards ratios were attenuated for both men and women. Model 5 shows no further attenuation of hazards ratios after additional adjustment for total cholesterol, diabetes, and smoking status. We observed similar patterns when we stratified the sample into white and black participants and repeated the analysis (data not shown).

DISCUSSION

Findings

When detected by the 12-lead ECG, left ventricular hypertrophy is known to carry a poor prognosis. We used competing Cox regression models to determine that men and women in the community with ECG LVH were more likely to experience any CVD event before non-CVD death over 15 years of follow up, independently of other cardiovascular risk factors. Our analysis of competing cumulative incidences and hazards ratios demonstrates that coronary heart disease/nonfatal myocardial infarction was the most likely first event in men with ECG LVH. In women with ECG LVH, the first event was more likely to be heart failure, followed closely by coronary heart disease/nonfatal myocardial infarction.

Implications

To the best of our knowledge, survival analysis in a competing risk framework has not been applied to subjects with ECG LVH. This method of analysis may reflect a more “real-world” approach, where individuals are at risk for multiple competing disease states simultaneously, which are associated with varying degrees of morbidity. Knowledge of the most likely first event, whether cardiovascular or non-cardiovascular, could inform primary prevention efforts and facilitate risk communication to patients. We found a somewhat different pattern of first events compared to a similar analysis of incident hypertensive individuals.[17] Although coronary heart disease remained the most common first event in men, heart failure was the most common first event in women in our cohort; in the previous study, stroke was the most common first event in hypertensive women. These results support the idea that LVH represents an intermediate phenotype in the progression of hypertensive heart disease, which may transition to cardiac failure.[3]

Cardiac dyssynchrony has been recognized as an increasingly important component in the pathophysiology of heart failure and was recently found to be associated with increased left ventricular mass and volume.[18] In our study, QRS duration was greater in men with ECG LVH compared to women with ECG LVH. However, women were more likely to experience heart failure as a first event. Potential explanations for this finding include a well-described “lag” of 10 years in CHD-related mortality in women relative to men;[13–14] the lower incidence of CHD in women in the age group in our analysis may partly account for the greater incidence of heart failure as a first event.

We confirmed the finding that individuals with ECG LVH were at significantly higher risk of any cardiovascular events compared to those without ECG LVH, independent of other established risk factors for CVD. The prevalence of ECG LVH was higher in blacks, who also have greater competing risks for non-CVD death in middle age; thus, the finding that cardiovascular events are more likely to occur first was not a foregone conclusion. Given the high prevalence of ECG LVH among black men and women in this cohort, we suspected that race was a key confounder in the analysis, and adjustment for age and race attenuated the hazards ratios, but the results remained significant. The other important potential confounder was systolic blood pressure and the hazards ratio for CVD event after adjustment for age, race, and systolic blood pressure remained significant. Possible mechanisms for increased risk of cardiovascular disease in patients with ECG LVH are complex and likely include myocardial ischemia, diastolic dysfunction, and arrhythmia, as well as the associated burden of risk factors.[19]

Our results for middle-aged women are particularly notable. The Third Report of the National Cholesterol Education Program’s Adult Treatment Panel (ATP-III) is one of the most common tools used in cardiovascular risk assessment.[20] However, the ATP-III algorithm has well-described limitations, notably that women are nearly always classified as low-risk.[21] We found that the cumulative incidence of CVD events in women with ECG LVH was 26.3% over 15 years, suggesting that women with ECG LVH are at particularly elevated risk beyond that associated with established risk factors in the Framingham risk score,[22] and should be considered for more intensive primary prevention efforts. Given the high risks for the blood pressure-related outcomes of heart failure and stroke and the documented benefits of blood pressure lowering for patients with ECG LVH,[23] attempts to lower blood pressure and regress ECG LVH may be particularly warranted.

Subclinical evidence of end-organ damage may be an effective method of determining the detrimental effects of a single uncontrolled risk factor. Regression of electrocardiographic LVH was found to reduce the risk of cardiovascular events in the Framingham Heart Study,[24] in a retrospective analysis. The Losartan Intervention for Endpoint Reduction in Hypertension (LIFE) study prospectively demonstrated that regression of ECG LVH reduced the risk of cardiovascular disease (CVD) endpoints, including cardiovascular death, nonfatal MI, and stroke.[23,25–26] The benefit was greater in those treated with losartan and part of the overall effect was independent of the reduction in blood pressure. Regression of ECG LVH by the angiotensin-converting enzyme inhibitor ramipril has also been shown to reduce cardiovascular risk in the Heart Outcome Prevention Evaluation (HOPE).[27]

Limitations of our study include the absence of QRS voltage-duration products, which may increase the sensitivity of the ECG to detect LVH, particularly the Cornell voltage-duration product.[28] However, the Sokolow-Lyon method has been well characterized, is easy to use in clinical practice, and was also used to define LVH in the HOPE and LIFE trials.[23,28] Cardiac imaging with echocardiography or magnetic resonance is more sensitive for detecting LVH;[29–30] however, ECG remains the first-line screening tool for diagnosis of LVH due to widespread availability, low cost, and simplicity. Despite these limitations, ECGs are commonly obtained in clinical practice for a variety of indications, and our study provides important prognostic information in men and women who are found to have ECG LVH. Another limitation is that subjects were classified based on a single baseline examination, and subsequent development of ECG LVH during the follow up period is not considered.

Conclusions

In individuals in the Atherosclerosis Risk in Communities study, middle-aged men and women with ECG LVH had a significantly higher risk of cardiovascular events over 15 years of follow up than those without ECG LVH. Individuals with ECG LVH are more likely to have a CVD event occur first before non-CVD death, and types of first CVD events are more likely to be CHD in men and HF in women, independent of the effect of other risk factors including hypertension. These results support the use of ECG LVH as an important risk marker in the epidemiology and prevention of cardiovascular disease.