Analogous to rapid ventricular pacing, frequent ventricular premature complexes (VPCs) can predispose over time to cardiomyopathy and subsequent heart failure (HF). We examined the association of frequent VPCs with HF incidence in a population-based cohort, free of HF and coronary heart disease at baseline. At study baseline (1987 to 1989), ≥1 VPC on a 2-minute rhythm electrocardiographic strip was seen in 5.5% (739 of 13,486) of the middle-age (45 to 64 years old at baseline) white and black, men and women of the Atherosclerosis Risk In Communities cohort. Incident HF was defined as the first appearance of International Classification of Diseases code 428.x in the hospital discharge record or death certificate through 2005. During an average follow-up of 15.6 years, incident HF was seen in 10% the participants (19.4% of those with VPCs vs 9.4% of those without). The age-, race-, and gender-adjusted hazard ratio of HF for VPCs was 1.89 (95% confidence interval 1.59 to 2.24). After multivariable adjustment for potential confounders, the hazard ratio of HF for those with any VPC versus no VPC was 1.63 (95% confidence interval 1.36 to 1.96). After additional adjustment for incident coronary heart disease as a time-varying covariate, the hazard ratio was 1.71 (95% confidence interval 1.42 to 2.08). Those with a greater frequency of VPCs or complex VPCs had similar rates of HF compared to those with a single VPC and all had rates greater than those with no VPC. In conclusion, in this large population-based cohort, the presence of VPCs was associated with incident HF, independent of incident coronary heart disease.
The occurrence of systolic dysfunction and clinical heart failure (HF) with chronic tachyarrhythmia, although uncommon, has been well described. Whether frequent ventricular premature complexes (VPCs), a common, yet mostly asymptomatic, ectopic arrhythmia are associated with incident HF is unknown. VPCs are associated with traditional risk factors of coronary heart disease (CHD), a greater incidence of CHD, and cardiovascular death. Those with frequent VPCs have been shown to have a reduction in ejection fraction and poor diastolic function. Finally, a few case reports and case series have pointed to the existence of VPC-induced cardiomyopathy, including its reversal with VPC ablation. The present study examined whether the presence of VPCs on a 2-minute electrocardiographic (ECG) recording is associated with incident HF in a large population-based cohort, and whether this putative association is present, independent of incident CHD.
Methods
The Atherosclerosis Risk In Communities (ARIC) study enrolled 15,792 subjects aged 45 to 64 years from 4 United States communities: Forsyth County, North Carolina; Jackson, Mississippi; 7 northwestern suburbs of Minneapolis, Minnesota; and Washington County, Maryland (1987 to 1989). Black residents were oversampled as a part of the area-probability sampling of Forsyth County, North Carolina, and enrollment at the Jackson, Mississippi site was restricted to black residents. A complete description of the ARIC communities and the design of the study has been previously published (available from: http://www.cscc.unc.edu/aric/ ). The cohort participated in a baseline visit and 3 follow-up examinations, as well as annual telephone interviews. At study baseline, information on the socioeconomic indicators, medical history, family history, cardiovascular risk factors, serum chemistries, ECG findings, and medication use was collected.
For the present analysis, participants with prevalent HF (Gothenburg criteria 3 or intake of medications, such as digoxin or intravenous diuretics for HF, n = 1,039 ) or missing information about incident HF were excluded. In addition, participants with missing values for the 12-lead electrocardiogram or 3-lead 2-minute rhythm strip (n = 74) or those with cardiac rhythm disturbances such as Wolf-Parkinson-White syndrome, atrial fibrillation/flutter, wandering atrial pacemaker, supraventricular tachycardia, or non-sinus rhythm were excluded (n = 283). Participants with missing or invalid values for key covariates (n = 703) or who self-identified their race as neither white nor black (n = 48) were excluded. Finally, those with prevalent or missing CHD data at baseline (n = 1,120) were also excluded. Some observations met >1 exclusion criterion. The remaining 13,486 cohort participants were included in the present study analyses.
In subset analyses involving adjustment for left ventricular hypertrophy using Cornell voltage, those with a QRS duration of ≥120 ms (n = 542) were also excluded, leaving 12,944 participants for the present analysis.
The participants with incident HF were identified using the International Classification of Diseases (ICD) , 9th or 10th revision, from hospital discharges (428.x) and from death certificates (ICD-9 or ICD-10 codes 428 and I50). Incident HF was defined as the first HF hospitalization with a HF discharge code or HF as the underlying cause of death on a death certificate (n = 68). HF cases were identified from 3 triennial re-examinations through 1998, annual telephone interviews to identify all hospitalizations, community hospital surveillance for hospitalizations, and searches of the National Death Index registry. Cohort follow-up for the present analysis ended on December 31, 2005. The validity of the ICD code for the diagnosis of HF hospitalization was found to be acceptable in the ARIC HF surveillance study—similar to the use of the Framingham criteria for HF (Rosamond et al, poster presented at the American Heart Association 2009 scientific sessions).
The participants were asked to fast for 12 hours and to refrain from smoking and consuming caffeinated beverages for 4 hours before the examination. A supine 12-lead electrocardiogram at rest and a 2-minute, 3-lead (V 1 , II, and V 5 ) rhythm strip were obtained using the MAC PC10 personal cardiogram (Marquette Electronics, Milwaukee, Wisconsin). The ECG data processing, monitoring, and quality control have been previously reported. Rhythm strips were classified 3 times by independent trained coders for total supraventricular, ventricular complexes, ventricular runs, and bigeminy, trigeminy, and multiform complexes. Adjudication of disagreement was performed by the ECG center’s principal investigator or the coding supervisor. The presence of VPCs and heart rate were determined from the rhythm strip.
The primary exposure was the presence of any VPC on the 2-minute rhythm electrocardiogram. VPCs seen on a 2-minute ECG rhythm strip are highly correlated with high-frequency VPCs seen on the 24-hour recordings. VPCs were classified by the frequency of their occurrence on the 2-minute rhythm strip (i.e., single VPC, ≥2 VPCs, and complex VPCs). Complex VPCs were defined as the presence of consecutive VPCs, multiple morphologies of VPCs, or the presence of short-coupled VPCs on the 2-minute ECG strip.
A parsimonious set of potential confounders for the relation of VPCs to incident HF was chosen on the basis of a strong physiologic relation to VPCs.
The study protocol manuals on the ARIC web site provide detailed information about the determination of prevalent CHD (manual 3), hypertension (manual 11), diabetes (manual 10), educational attainment, medication intake history, and Cornell voltage estimated left ventricular mass in the ARIC study. Similarly, the definition of these covariates can be found in the variable dictionary at the ARIC study web site.
The time-to-event for those without incident HF was censored at death, loss to follow-up, or December 31, 2005, whichever occurred earlier. The rate of HF by race, gender, and VPC categories were estimated, assuming a constant rate over time and a Poisson distribution. Cox regression models were used to estimate the hazard ratio of HF, contrasting the presence of any VPC with no VPC. Finally, multivariate Cox regression models were fit, including interaction terms between main exposure and age categories, race, gender, diabetes, hypertension, and heart rate (categories at the 95th percentile), after adjusting for other potential confounders associated with prevalent VPCs and HF. Relations by increasing frequency or complexity of VPCs were examined using dummy variables for the defined categories. Furthermore, to examine the effect of a greater frequency of VPCs on HF, we also classified the frequencies as 0, 1, 2 to 4, and ≥5.
As additional analyses, the Cox model was adjusted for incident CHD as a time-varying covariate, and an adjustment for left ventricular mass using Cornell voltage was done in a subset analysis, because this required additional exclusion of some study participants. The proportional hazard assumption was evaluated using graphic methods [ln-ln S (t) graphs] and Cox tests and was not invalidated. Multivariate-adjusted cumulative HF curves were plotted by the presence of VPC using inverse probability weights.
All statistical computations were performed using SAS software, version 9.1.3 (SAS Institute, Cary, North Carolina). A p value of <0.05 for a 2-sided null hypothesis was considered statistically significant for the main covariates. A p value of <0.2 was considered statistically significant for the interaction terms.
Results
At least a single VPC was identified in 739 (5.5%) of the 13,486 study participants using the baseline rhythm strip recorded for a 2-minute duration. The differences in the characteristics of the study participants by the presence of any VPC and by incident HF status are listed in Table 1 .
| Characteristic | Any VPC on 2-min ECG Strip | HF During Follow-Up | ||
|---|---|---|---|---|
| Yes (n = 739) | No (n = 12,747) | Yes (n = 1,344) | No (n = 12,142) | |
| Age (years) | 56.0 ± 5.7 | 53.8 ± 5.7 | 56.5 ± 5.4 | 53.7 ± 5.7 |
| Women | 49.4% | 56.4% | 50.6% | 56.6% |
| Black race | 29.9% | 25.8% | 36.5% | 24.8% |
| Education | ||||
| Less than high school | 27.9% | 21.9% | 39.9% | 20.3% |
| High school, less than college | 41.0% | 41.2% | 36.8% | 41.7% |
| College or more | 31.1% | 36.9% | 23.4% | 38.0% |
| Body mass index (kg/m 2 ) | 28.2 ± 5.5 | 27.4 ± 5.2 | 29.6 ± 6.3 | 27.2 ± 5.1 |
| Low-density lipoprotein cholesterol (mg/dl) | 136.5 ± 38.3 | 137.1 ± 39.1 | 142.6 ± 40.2 | 136.5 ± 38.9 |
| High-density lipoprotein cholesterol (mg/dl) | 51.4 ± 16.6 | 52.6 ± 17.0 | 48.1 ± 15.7 | 53.0 ± 17.0 |
| Potassium (mmol/L) | 4.4 ± 0.5 | 4.4 ± 0.5 | 4.4 ± 0.5 | 4.4 ± 0.5 |
| Magnesium (mEq/L) | 1.6 ± 0.2 | 1.6 ± 0.2 | 1.6 ± 0.2 | 1.6 ± 0.2 |
| Heart rate (beats/min) | 68.0 ± 10.8 | 66.4 ± 10.1 | 69.1 ± 11.5 | 66.2 ± 9.9 |
| Cornell voltage ⁎ (mV) | 1.3 ± 0.6 | 1.2 ± 0.5 | 1.4 ± 0.6 | 1.2 ± 0.5 |
| Systolic blood pressure (mm Hg) | 124.8 ± 19.0 | 120.6 ± 18.5 | 130.6 ± 21.6 | 119.8 ± 17.9 |
| Hypertension | 41.5% | 30.9% | 54.0% | 29.0% |
| Antihypertensive agents | 29.9% | 21.2% | 39.6% | 19.7% |
| Glucose (mg/dL) | 109.5 ± 40.0 | 106.9 ± 36.8 | 130.5 ± 69.2 | 104.4 ± 30.4 |
| Diabetes | 11.5% | 8.1% | 25.7% | 6.3% |
| Current smoking | 25.2% | 25.9% | 37.6% | 24.5% |
| Former smoking | 33.6% | 31.3% | 31.0% | 31.5% |
| Pack-years of smoking | 17.3 ± 23.1 | 14.9 ± 20.6 | 22.7 ± 26.1 | 14.2 ± 19.9 |
| Alcohol intake (g/wk) | 45.9 ± 118.8 | 42.1 ± 92.1 | 38.7 ± 91.5 | 42.7 ± 94.0 |
| Any VPCs † | 100% | 0% | 10.6% | 4.9% |
| Single VPC | 40.3% | 0% | 3.9% | 2.0% |
| >1 VPC | 48.4% | 0% | 5.3% | 2.4% |
| Complex VPCs | 11.2% | 0% | 1.4% | 0.5% |
⁎ Cornell voltage was calculated in those with QRS <120 ms, with nonmissing values, and not having prevalent coronary heart disease (CHD).
† VPC coded by investigators using 2-min (3-lead ECG strips) at baseline; complex VPCs defined by presence of consecutive, multimorphic, or short-coupled (<5 percentile).
Incident HF was seen in 10% of subjects during an average ± SD follow-up period of 15.6 ± 3.8 years. The cumulative proportion of incident HF was greater (19.4%) in those with any VPCs than in those without (9.4%). The unadjusted rate of HF in those with any VPCs was more than twice that of those without (rate/1,000 person year 13.5 vs 6.0, p < 0.001).
The age-, race-, and gender-adjusted hazard ratio (HR) of HF for any VPCs was 1.89 (95% confidence interval [CI] 1.59 to 2.24). After multivariate adjustment for age, gender, race, study center, education level, blood glucose level, diabetes, systolic blood pressure, intake of antihypertensive medications, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, body mass index, smoking status, pack-years of smoking, alcohol intake, heart rate, serum potassium ion, serum magnesium ion, the HR of HF among those with any VPCs versus no VPCs was 1.63 (95% CI 1.36 to 1.96). Figure 1 shows the multivariate-adjusted cumulative proportion of HF by the presence of any VPCs.
These results did not change appreciably after additional adjustment for left ventricular mass using Cornell voltage (HR 1.59, 95% CI 1.32 to 1.92) or medications such as β blockers, anti-arrhythmic agents, and calcium channel blockers. Also, the association existed after adjustment for incident CHD as a time-varying covariate (HR 1.71, 95% CI 1.42 to 2.08).
The rate of HF was similar in the participants with either a greater frequency or greater complexity of VPCs compared to those with a single VPC ( Figure 2 ). Also, all these participants had greater rates than those without any VPC. However, when examining the effect of a greater frequency as 1, 2 to 4, or ≥5 compared to none, the HR was 1.84 (95% CI 1.41 to 2.41), 1.98 (95% CI 1.49 to 2.63), and 2.16 (95% CI 1.56 to 3.00), respectively.
