Background Premature ventricular contractions (PVCs) are associated with an increased risk of morbidity and mortality. Therefore, it was aimed to assess risk factors for the frequency of PVCs in young and healthy adults.
Methods Our population-based study included 2048 healthy adults from the general population aged 25–41 years. PVC frequency was determined by 24-hour Holter ECG. We performed multivariable regression analysis using stepwise backward selection to identify factors independently associated with PVC frequency.
Results Median age was 37 years, 953 (46.5%) were male. At least one PVC during the 24-hour monitoring period was observed in 69% of participants. Median number of detected PVCs was 2, the 95th percentile was 193. In multivariable regression analyses, we found 17 significant risk factors for PVC frequency. Low educational status (risk ratio (RR) 3.33; 95% CI 1.98 to 5.60), body height>median (1.58, 95% CI 1.11 to 2.24) and increasing levels of waist:hip ratio (2.15, 95% CI 1.77 to 2.61), N-terminal pro brain natriuretic peptide (1.52, 95% CI 1.30 to 1.76) and Sokolow-Lyon Index (1.38, 95% CI 1.15 to 1.66) (all p≤0.01) were associated with a higher PVC frequency. Physical activity (RR fourth vs first quartile 0.51, 95% CI 0.34 to 0.76) and increasing levels of haemoglobin (0.58, 95% CI 0.47 to 0.70) and glucagon-like peptide-1 (0.72, 95% CI 0.64 to 0.82) (all p<0.001) were related to a lower PVC frequency.
Conclusions PVC occurrence is common even in healthy low-risk individuals, and its frequency is associated with several covariates mainly related to cardiovascular risk factors, markers of cardiac structure and function and socioeconomic status.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Several studies have established that individuals with premature ventricular contractions (PVCs) have a higher risk of ischaemic heart disease,1 stroke,2 ,3 ventricular dysfunction,4 sudden cardiac death5 ,6 and all-cause mortality.1 ,6 ,7 For example, a meta-analysis of studies in adults without clinically apparent heart disease demonstrated an OR for the combined endpoints of all-cause mortality, cardiovascular mortality, sudden cardiac death or the development of ischaemic heart disease of 1.72 (95% CI 1.28 to 2.31) in subjects with documented PVCs compared with those without.6 This increased risk is evident in individuals with8 ,9 and without4–6 prevalent cardiovascular disease. At least some of those associations seem to be causal, as PVC removal reversed functional and structural changes.10–12
Due to these potential adverse risks of PVCs, an enhanced understanding of their causes and risk factors is of major clinical importance. However, little information is available on frequency and risk factors of PVCs in the general population. We therefore evaluated the relationships of a broad set of characteristics and biomarkers with PVC count in a large and well-characterised cohort of young and healthy adults.
All inhabitants of the Principality of Liechtenstein aged between 25 and 41 years were asked to participate in the “Genetic and phenotypic determinants of blood pressure and other cardiovascular risk factors” (GAPP) study. The exact study methodology has been published previously.13 Main exclusion criteria were an established cardiovascular disease, known obstructive sleep apnoea, renal failure, current intake of antidiabetic drugs, other severe comorbidities or a body mass index (BMI) >35 kg/m2. Of the 2170 GAPP participants, we excluded 122 individuals due to missing 24-hour Holter ECG (24-hour ECG) recording (n=16), 24-hour ECG duration <80% (n=22), missing or incomplete blood samples (n=18), missing resting ECG (n=10) and incomplete questionnaires (n=56). Thus, 2048 (94.4%) participants with complete data remained for the current analysis. The study protocol was approved by the local ethics committee and written informed consent was obtained from each participant.
Assessment of PVCs
In every participant a 24-hour ECG was obtained using a validated three channel device (AR12plus, Schiller AG, Switzerland). If the monitoring time was <80% of the target time (ie, <19.2 hours) or recording quality was low, the 24-hour ECG was repeated if possible. Postprocessing was conducted by a trained research associate under the supervision of a cardiologist and using a dedicated software (Medilog Darwin V2, Schiller AG, Switzerland). All artefacts were completely removed and PVCs were identified. PVCs were defined as premature beats with a QRS duration of >120 ms and different QRS-wave and T-wave morphology compared with normal sinus beats. In case of ambiguity an experienced electrophysiologist was consulted.
Assessment of laboratory parameters
Fasting venous blood samples were obtained from every participant by venipuncture, as described previously.13 Plasma levels of potassium, sodium, calcium, creatinine, aspartate transaminase, low density lipoprotein cholesterol (LDL-C), high density lipoprotein cholesterol (HDL-C), high sensitivity C reactive protein (hs-CRP) and N-terminal pro brain natriuretic peptide (NT-proBNP) were quantified on a Roche Cobas 6000 analyser (F. Hoffmann-La Roche, Switzerland). Haemoglobin (Hb) and complete blood count was determined using the sodium-lauryl-sulfate haemoglobin method (Sysmex SE 5000). Glycated haemoglobin (HbA1c) was obtained using high performance liquid chromatography (Bio-Rad Laboratories AG, Switzerland). Plasma copeptin levels were measured using a Brahms Luminometry analyser. Plasma levels of high sensitivity cardiac troponin I (hs-cTnI), endothelin-1 and glucagon-like peptide-1 (GLP-1) were measured using a single-molecule counting technology (Erenna Immunoassay System, Singulex, Alameda, California, USA) from frozen plasma samples stored at −80°C.
Individuals with undetectable hs-CRP (n=28) or hs-cTnI (n=11) levels were assigned a value of 0.15 mg/L or 0.04 ng/L, respectively. We calculated the neutrophil:lymphocyte ratio by dividing the total number of neutrophils by the total number of lymphocytes. Dysglycemia was defined as HbA1c levels of 5.7% or higher.
Assessment of other study variables
Personal, medical and nutritional factors were assessed using standardised questionnaires. Smoking status was dichotomised into current and non-current smokers (past and never smokers). Among current smokers, pack-years of smoking were calculated by multiplying the number of years smoked with the average number of cigarettes smoked per day. Highest educational status achieved was categorised into high school degree, college degree or university degree. Physical activity was evaluated with the validated individual physical activity questionnaire. Total amount of vigorous physical activity per week was calculated and quartiles were used for all analyses. Regular fruit and vegetable consumption was defined as consuming at least five servings of fruits and/or vegetables per day. Weight and height were directly measured using standardised devices. BMI was calculated by dividing weight in kilograms by height in metres squared. Waist circumference was obtained at the tightest part between the costal arch and iliac crest, hip circumference was measured round the widest portion of the buttocks. Waist:hip ratio was defined as the ratio of these parameters.
Three conventional blood pressure (BP) measurements were obtained in a sitting position after 5 min of rest using a validated oscillometric device (Microlife BP3AG1; Microlife AG, Switzerland).13 The mean of the second and third reading was used for our analysis. Standard 12-lead ECGs were obtained in all study participants and the resting heart rate was calculated using the mean RR interval. The Sokolow-Lyon Index (SLI) was calculated as the sum of the higher S-wave in lead V1 or V2 and the higher R-wave in lead V5 or V6.
Baseline characteristics were stratified using the 25th, 50th, 75th and 90th percentile of the PVC count as cut-offs. Distribution of continuous variables was checked using skewness, kurtosis and visual evaluation of the histogram. Normally distributed continuous variables were displayed as mean (SD) and otherwise as median (IQR). Group comparisons were conducted using one-way analysis of variance or Kruskal-Wallis tests, respectively. Categorical variables were presented as numbers (percentages) and compared using χ2 tests.
Negative binomial regression models using PVC frequency as the outcome variable were constructed to identify independent risk factors of PVC count in the population. The linearity of the association for continuous variables was checked using risk factor specific quartiles in sex and age adjusted regression models. Variables showing approximate linear relationships with PVC count were entered in the models as continuous variables. Otherwise, quartiles were entered in the regression models. Skewed variables with an approximate log-norm distribution were log-transformed.
We then applied stepwise backward selection with a p value for staying in the model of 0.05. The variables listed in figure 1 and additionally sex, dysglycemia, systolic office BP, BMI, QRS duration, serum calcium, potassium, LDL-C, HDL-C, hs-CRP and hs-cTnI, were considered as candidates for entering the model. Given the important overall influence of age and sex on the occurrence of cardiovascular events, these two variables were forced in the regression models independent of the significance of their association with PVCs. To allow a better comparability between continuous variables, risk ratios (RRs) were calculated per one SD increase. We checked for multicollinearity in our models using correlation matrix and variance inflation factor, and found no evidence for it. In a separate step, univariate and multivariable logistic regression analysis were performed using the 90th percentile of the PVC distribution as a cut-off for the outcome variable. Age, sex and all covariates that were significantly associated with PVC count in the negative binomial regression model were included in this multivariable logistic regression model. SAS V.9.4 (Cary, North Carolina, USA) and STATA V.13.1 (College Station, Texas, USA) statistical software packages were used to conduct the analyses. A p value <0.05 was considered to indicate statistical significance.
Stratified baseline characteristics are shown in table 1. There were significant between group differences for age, sex, low educational status, creatinine and NT-proBNP. Of the 2048 participants included in this analysis, 1407 (68.7%) had at least one PVC during their 24-hour ECG. The recording of the Holter ECG was 24 hours in 2036 (99.4%) participants. The median number of PVCs was 2, the 95th percentile and the 99th percentile were 193 and 2545, respectively. Figure 2 shows the distribution of PVC frequency. Overall, 1080 subjects (52.7%) performed at least 150 min of moderate or vigorous physical activity per week, and 25% of participants performed at least 360 min of physical activity per week.
In multivariable stepwise backward regression analysis 17 independent risk factors for PVC count were identified and are presented in figure 1. Overall, 10 parameters were associated with an increased risk for PVCs, while seven showed an inverse relationship with PVC frequency. Low educational status having smoked ≥15 pack-years, increasing waist:hip ratio, regular fruit and vegetable consumption, height>median, increasing levels of NT-proBNP and increasing SLI were associated with a higher PVC frequency. Plasma levels of GLP-1, and physical activity were associated with a lower PVC frequency. In the final model, both age (per SD) (1.15 (0.99; 1.32), p=0.06) and sex (male sex 1.46 (0.79; 2.68), p=0.22) were not related to PVC count.
The results of the univariate and multivariable logistic regression using the 90th percentile of the PVC distribution (ie, 32 PVCs per 24 hours) as the outcome variable were similar to the main analysis and are presented in table 2. The OR (95% CI) from the multivariable regression model for having a PVC frequency over the 90th percentile was 1.29 (1.10; 1.52), p=0.003 per one SD increase in age, 1.28 (1.07; 1.54), p=0.008 per one SD increase in NT-proBNP and 1.52 (1.01; 2.28), p=0.04 for body height over the median.
In this large population-based study of well-characterised young and healthy adults, the majority of participants had at least one PVC during 24-hour Holter ECG monitoring. This is an important finding, as prior studies have shown that the occurrence of PVCs is associated with an increased risk of cardiovascular morbidity and mortality even in subjects without known cardiovascular disease and that at least some of these associations may be causal.5 ,6 In the current study, we were able to describe the PVC frequency and also to define a large number of associated risk factors, the most important being a low educational status, high waist:hip ratio, high amount of pack-years, high NT-proBNP levels, high SLI, higher age, physical inactivity, low haemoglobin and plasma levels of GLP-1. On the other hand, in our study only few individuals had a very high PVC count and the 90th percentile was 32 PVCs over 24 hours. Because different ECG recording lengths were used in different studies, it is difficult to compare the PVC frequency across different studies.3 Compared with the PVC frequency in this young population, a slightly higher PVC frequency (median PVC count of 9) was found among an older population with various comorbidities.4 Future studies need to assess whether there is a truly linear association of the PVC count with adverse outcomes, or whether there is a threshold effect, especially at the lower end of the PVC count spectrum.
We are aware of only one large prior study that assessed predictors of PVCs. The strongest predictors for PVCs in this study were age, male sex, African-American ethnicity, low educational attainment, hypertension, organic heart disease, plasma magnesium levels and sinus rate. However, it has to be emphasised that the PVC prevalence in that previous study was only around 6%, as it used 2 min ECG recordings to quantify PVCs.14 Not surprisingly, the length of ECG monitoring has previously been shown to be a key determinant in PVC quantification. Thus, our study provides important additional insights, as we were also able to evaluate risk factors for overall PVC count, which by itself is an important prognostic determinant.
The relationship of increased ventricular mass and ventricular arrhythmias has been described previously.15 For instance, an analysis of the Framingham Heart Study showed that men, but not women, with left ventricular hypertrophy defined by ECG criteria were at higher risk for ventricular arrhythmias.16 Our study confirms and extends this finding as we found a continuous association of left ventricular mass assessed by SLI and PVC frequency, irrespective of sex. Of note, left ventricular hypertrophy increases vulnerability to electrical stimulation,17 such that left ventricular mass may play a crucial role in the occurrence of PVCs. This association would probably have been even stronger if cardiac imaging had been available in our study, given the relatively low sensitivity of ECG-based criteria to quantify left ventricular mass and hypertrophy.16
Despite the known increase of left ventricular mass in subjects doing high amounts of physical activity, physical activity was associated with a decreased PVC count in our study. This finding suggests protective effects of exercise that may outweigh potential proarrhythmic effects of an exercise-induced increase of ventricular mass. Possible protective mechanisms of physical activity could be reduction of abdominal adiposity, sympathetic nervous tone and oxidative stress.18 It is well known that higher levels of exercise lead to a lower resting heart rate by modulating autonomic nervous tone. Our finding that an increased resting heart rate is a risk factor for PVC count is in line with the exercise-heart rate relationship and supports the hypothesis that a higher sympathetic nervous tone predisposes to PVC occurrence.
In this context, adult height is also a major determinant of heart size and chamber dimensions.19 Prior studies have shown that height is a strong and independent predictor for the occurrence of premature atrial contractions20 and incident atrial fibrillation.21 Our study now suggests that height is also a major risk factor of ventricular arrhythmias. Whether this is due to increased chamber dimensions or other mechanisms needs to be defined in future studies.
Prior studies have shown that left ventricular stretch induces depolarisations that trigger arrhythmias in an experimental animal model.22 In addition, left ventricular hypertrophy occurs as a response to increased left ventricular wall stress.23 In line with this evidence we found several indirect markers of left ventricular stretch to be associated with an increased PVC count, including plasma levels of NT-proBNP, and hypervolemia-related markers such as low copeptin levels, sodium levels or low haemoglobin concentration.24 While none of the participants in this cohort is expected to have manifest hypervolemia, subclinical hypervolemia may still be associated with PVC count. In addition, obesity is also a major determinant of cardiac chamber size, wall stress, hypervolemia25 and sympathetic nervous activity26 providing many potential mechanisms why obesity was strongly associated with a higher PVC burden. Finally, BP was not an independent risk factor of PVCs in the current study, despite being a major determinant of left ventricular mass and wall stress, and although an association with PVCs has been reported previously.14 ,27 A possible explanation is that the prevalence of hypertension in the current study was relatively low and that in this cohort of young and healthy adults it has not been present long enough to manifest its detrimental impacts on heart rhythm.
A low socioeconomic status (SES) has been repetitively related to an increased risk of cardiovascular risk factors28 and the occurrence of cardiovascular events.29 People with a low SES tend to have a less healthy lifestyle and a higher prevalence of obesity. Many of the detrimental lifestyle and metabolic factors associated with a low SES were independent risk factors of PVC occurrence in this study, namely heavy smoking, high waist:hip ratio, low GLP-1 levels and physical inactivity. Interestingly, a low educational status remained a strong risk factor for PVC occurrence and frequency even after adjusting for these factors, suggesting that a low SES has other effects promoting the occurrence of PVCs, for example, psychological stress.30 In addition, a higher PVC burden may be one additional factor why individuals with a low SES have a higher incidence of cardiovascular events.
Unexpectedly, consuming elevated amounts of fruits and vegetables was associated with a higher PVC count. A possible explanation for this could be inverse causation. Subjects with a high burden of cardiovascular risk factors or symptomatic PVCs may be more likely to engage in healthy eating habits.
Strength and limitations
Strengths of our study include the large sample size of well-characterised healthy individuals, the population-based design and the use of 24-hour ECG to quantify PVCs. Possible limitations which need to be taken into account in the interpretation of the results include the following: first, we did not have cardiac imaging data for our analysis and therefore had to use SLI as an indirect marker of left ventricular hypertrophy. Second, the cross-sectional design does not allow to draw conclusions about the directionality of the observed associations. Third, our study population dominantly consists of healthy Caucasian adults and the generalisability of these results to other cohorts such as athletes or patients suspected to have structural heart disease remains unclear.
In this large population-based study we found that PVC occurrence is common in young and healthy low-risk individuals. The PVC count is associated with a large number of cardiovascular risk factors, markers of cardiac structure and function and SES. Future studies are needed to assess whether modifying these factors will help to lower the PVC count and help to improve cardiovascular outcomes in the population.
What is already known on this subject?
Premature ventricular contraction count is an important predictor for the development of heart failure and cardiovascular mortality.
What might this study add?
In young and healthy adults from the general population, premature ventricular contraction count is associated with traditional cardiovascular risk factors, markers of cardiac structure and function and the educational status.
How might this impact on clinical practice?
Patients presenting with a large number of risk factors for premature ventricular contractions (PVCs) might be monitored for frequent PVCs and subsequent damages. However, additional research is needed to determine whether modifying these risk factors will reduce the population PVC burden.
Contributors Data collection: MvR, SA, MB, SB and SS. Statistical analysis: MvR, SA, TS and DC. Principal investigators: DC, LR and MR. Analysis biomarkers: JT and JE. Draft manuscript: MvR and DC. Careful editing of the manuscript: SA, MB, TS, SB, SS, JE, JT, MR, LR, DC and MvR.
Funding The GAPP study was supported by the Liechtenstein Government, the Swiss Heart Foundation, the Swiss Society of Hypertension, the University of Basel, the University Hospital Basel, the Hanela Foundation, Schiller AG and Novartis. DC was supported by a grant of the Swiss National Science Foundation (PP00P3_133681 and PP00P3_159322). MB was supported by a grant of the University of Basel and Freiwillige Akademische Gesellschaft (FAG) Basel. Plasma levels of high sensitivity cardiac troponin I, endothelin-1 and glucagon-like peptide-1 were determined free of charge by Singulex.
Competing interests JE and JT are employees of Singulex.
Patient consent Obtained.
Ethics approval Ethikkommission Zürich, Switzerland.
Provenance and peer review Not commissioned; externally peer reviewed.