Heritability and causality of QRS duration and chronic heart failure risk

Abstract

Aims

Observational studies report conflicting results on the relationship between QRS duration and chronic heart failure (CHF), presenting challenges in establishing a causal link. This study investigates the heritability of QRS duration and CHF and their causal relationship.

Methods and results

Genome‐wide association studies (GWAS) cohort for QRS duration included 10 815 samples from the IEU Open GWAS project, while exome‐wide association studies (EWAS) data were sourced from the CHARGE Exome‐Chip EKG consortium, involving 77 898 European individuals. The CHF GWAS dataset comprised 486 160 samples from the EMBL‐EBI GWAS catalogue. Heritability estimates were determined using linkage disequilibrium score regression (LDSC). Mendelian randomization (MR) and sensitivity analyses assessed the causality. Heritability estimates for QRS duration were 16.3% from GWAS and 18.5% from EWAS. CHF exhibited minimal genetic influence with a heritability estimate of 0.8%. Six variants from the GWAS and 27 variants from the EWAS, including those in ion channel‐related genes, like CASQ2, SCN5A and SCN10A, were identified as instrumental variables. MR analysis indicated that shorter QRS duration is causally associated with an increased CHF risk [QRS GWAS: (IVW (MRE): OR 0.84, 95% CI 0.78–0.91, P = 2.26E‐05); QRS EWAS: (IVW (MRE): OR 0.98, 95% CI 0.96–0.99, P = 6.57E‐05)]. Sensitivity analyses confirmed the robustness of these findings [corrected GWAS: Egger_intercept = 0.002, P = 0.94; corrected EWAS: Egger_intercept = 0.007, P = 0.38].

Conclusions

This study establishes a causal relationship between shorter QRS duration and increased CHF risk, highlighting the importance of genetic factors in cardiac electrical conduction. Identifying QRS duration as a genetic marker for CHF risk can enhance early diagnosis and personalized treatment strategies.

Methodological framework used to investigate the heritability using linkage disequilibrium score regression (LDSC) and causal relationship between QRS duration (exposure) and chronic heart failure (outcome) using Mendelian randomization. QRS data were sourced from GWAS and exome‐wide association studies (EWAS), and chronic heart failure GWAS data were derived from the NHGRI‐EBI catalogue. Genetic proxies (single nucleotide polymorphisms) as instrumental variables were selected based on core MR assumptions, including relevance, independence and exclusion of confounders. Mendelian randomization analyses were used to infer causality, with sensitivity tests ensuring the robustness of the results. Black arrows represent the analysis process, and red arrows indicate MR core assumptions.

1. Introduction

Cardiovascular diseases (CVDs) continue to be a primary cause of morbidity and mortality worldwide, with chronic heart failure (CHF) representing a significant component of this burden. ¹ Understanding the genetic basis of CHF is critical for developing targeted interventions and improving patient outcomes. ² One potential genetic marker of interest is QRS duration (normal <120 ms), which measures the time taken for ventricular depolarization and can indicate cardiac electrical conduction abnormalities.

Recent genetic association studies have indicated that a considerable portion of the variance in QRS duration can be attributed to genetic factors, with heritability estimates ranging from negligible to 36%–43%. ³ , ⁴ , ⁵ Observational studies have established QRS duration as a risk factor for heart failure, with changes in QRS duration from baseline to 3 months being independently associated with long‐term survival and left ventricular ejection fraction. ⁶ The Framingham Heart Study found that each standard deviation increase in log‐transformed QRS duration was associated with a 23% higher risk of CHF. ⁷ Prolonged QRS duration is prevalent in CHF patients and correlates with a poorer prognosis. ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² Some studies have demonstrated that patients with prolonged QRS duration experience less left ventricular reverse remodelling compared to those with narrow QRS duration, without cardiac resynchronization therapy (CRT). On the other hand, in patients with a QRS duration of less than 130 ms and left ventricular dyssynchrony, particularly in males, the mortality rate after receiving cardiac resynchronization therapy (CRT) is significantly higher, exceeding 81%. Compared with those with prolonged QRS duration (<160 ms), patients with short QRS who receive CRT exhibit poorer improvements in left ventricular function and reverse remodelling. ¹² , ¹³ , ¹⁴ , ¹⁵ These findings suggest that a narrow QRS duration may also imply a worse prognosis. Moreover, the extent to which this genetic variation in QRS duration influences the risk of developing CHF remains unclear.

Previous research has primarily focused on the prognostic value of QRS duration in existing heart failure patients and has reported inconsistent results, without exploring its causal role in the development of CHF. Additionally, observational studies cannot fully exclude the influence of reverse causation and confounding factors, making it challenging to establish a true causal relationship. ¹⁶ This gap in knowledge underscores the necessity of investigating the causal relationship between QRS duration and CHF.

This study aims to utilize large‐scale genome‐wide association study (GWAS) and exome‐wide association study (EWAS) data to assess the extent to which genetic components can explain QRS duration and CHF through linkage disequilibrium score regression (LDSC). ⁴ , ¹⁷ Furthermore, Mendelian randomization (MR) will be employed to confirm the causal relationship between QRS duration and CHF. The advantage of MR lies in its ability to mimic the principles of randomized clinical trials (RCTs) by using genetic allele allocation to exclude reverse causation and real‐world confounding factors, employing genetic variants of the exposure as instrumental variables for causal analysis. ¹⁶ , ¹⁸ By confirming the causal effect of genetically determined QRS duration on CHF, we hope to improve the clinical diagnosis and prognostic management of high‐risk heart failure patients and potentially pave the way for new therapeutic strategies.

2. Methods

2.1. Data sources

Genetic data representing QRS duration and CHF were retrieved and downloaded from the NHGRI‐EBI GWAS catalogue (https://www.ebi.ac.uk/) and the IEU Open GWAS project (https://gwas.mrcieu.ac.uk/). The GWAS cohort for QRS duration from the Neale lab included 10 815 samples with 13 479 245 variants (ID: ukb‐d‐12340_irnt), while QRS duration EWAS data were sourced from the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Exome‐Chip EKG consortium, involving a meta‐analysis of 77 898 European individuals (ID: GCST007103). A total of 228 164 polymorphic genes on the exome‐chip array passed quality control and were included in the analysis. QRS duration was measured in milliseconds (ms). Exclusions were applied for individuals with a QRS duration >120 ms, as prolonged QRS intervals are commonly observed in various pathological conditions, such as myocardial ischaemia, ventricular hypertrophy and cardiac conduction system disorders. These conditions can confound the interpretation of genetic background studies. Additional exclusion criteria included atrial fibrillation (AF) on baseline electrocardiogram, a history of myocardial infarction or heart failure, Wolff–Parkinson–White syndrome (WPW), pacemaker implantation or the use of Class I and III antiarrhythmic medications (ATC classification C01B*). Further quality control details are available in the original study. ⁴ CHF GWAS data were sourced from a study published in Nature Genetics, involving a cross‐phenotype association study with 486 160 samples and 24 178 220 single nucleotide polymorphisms (SNPs). ¹⁷ It was obtained from NHGRI‐EBI catalogue through ID: GCST90018806. The Genome Reference Consortium Human Build 37 (GRCh37) /hg19 was used. Ethical approvals were obtained in the respective original studies, and re‐analysis of these data did not require additional ethical approval.

2.2. Genetic instruments of QRS duration used for causal inference

SNPs used as proxies for QRS duration must satisfy three core MR assumptions: relevance, independence and exclusion restriction. A significance threshold of P < 5 × 10⁻⁸ was set to ensure strong relevance, with an F‐statistic threshold of 10 ((beta/SE)²) applied to filter out weak instruments. Independence was ensured by linkage disequilibrium (LD) clumping (r ² = 0.001, window = 10 Mb). The LDtrait module in NCBI LDlink was used for SNP‐phenotype screening to exclude potential confounders (BMI, smoking and alcohol consumption, hypertension, diabetes and ischaemic cardiomyopathy) to ensure exclusive causal pathways. Selected variants were harmonized with the outcome GWAS for removing palindromic variants with allele frequencies close to 0.5.

2.3. Statistical analyses

The heritability of QRS duration and CHF traits was estimated using LD regression scores (LDSC) R package (version 0.1.0). Causal effects and sensitivity analyses were conducted using the TwoSampleMR package (version 0.6.2) in R. Radial MR (version 1.1) and mr.raps (version 0.4.1) were used for detection and correction of pleiotropic outliers and diagnostic plotting of MR assumptions, respectively. Causal effects were primarily estimated using inverse variance weighted (multiplicative random effects) [IVW (MRE)] method, assuming all SNPs as valid instruments, and by regressing the SNP‐CHF effect on the SNP‐QRS duration effect, weighted by the inverse of the SNP‐CHF standard error. Robust analyses were conducted using MR Egger regression, weighted median model, IVW radial and robust adjusted profile score (RAPS). Positive results were considered significant if IVW (MRE) and any robust method were significant. Sensitivity analyses included Egger intercept tests for horizontal pleiotropy and leave‐one‐out tests for single SNP influence. Potential pleiotropic variants were identified and adjusted using Radial MR and re‐evaluated for pleiotropy and normality through RAPS. Positive causal inference thresholds P of 0.05 were considered robust if the causal estimate passed all sensitivity analyses (pleiotropic P > 0.05).

3. Results

3.1. Heritability of phenotypes assessed by LDSC

Using LDSC, we assessed the heritability of QRS duration and chronic CHF. For QRS duration, the GWAS data indicated a heritability estimate (h ²) of 0.163 (SE = 0.058), with 16.3% of the variance explained by genetic factors (Z‐score: 2.810, P‐value: 5.01E‐03). The EWAS data provided a heritability estimate of 0.185 (SE = 0.037), showing that 18.5% of the variance is genetically determined (Z‐score: 5.001, P‐value: 5.70E‐07). In contrast, CHF GWAS data showed a much lower heritability estimate of 0.008 (SE = 0.001), indicating that only 0.8% of the variance is due to genetic factors (Z‐score: 7.830, P‐value: 4.78E‐15) (Table  1 ). In contrast, CHF exhibits minimal genetic influence, highlighting the significant role of environmental and lifestyle factors in its development.

Table 1.

Heritability of phenotypes assessed by LDSC

Phenotypes	h2	h2_Standard error	h2_Z	h2_P value
QRS GWAS	0.163	0.058	2.810	5.01E‐03
QRS EWAS	0.185	0.037	5.001	5.70E‐07
CHF GWAS	0.008	0.001	7.830	4.78E‐15

h², heritability estimate, indicating the proportion of variance in the phenotype that can be attributed to genetic factors; h² Standard error, the standard error of the heritability estimate, reflecting the uncertainty in the estimate; h² Z, the z‐score for the heritability estimate, a measure of the statistical significance; h² P‐value, the P‐value associated with the z‐score, indicating the strength of evidence against the null hypothesis.

3.2. Genetic tools for QRS duration

GWAS and identified leading phenotypic variants for QRS duration, providing insights into cardiac electrical conduction. To investigate whether QRS duration genetically determines the risk of CHF, we identified eligible SNPs as genetic instruments for causal inference. Six SNPs were identified from the GWAS with F‐statistics ranging from 31.64 to 65.60. To address this, we included a larger sample size of EWAS data focusing on coding genes and identified 27 additional SNPs as instruments with F‐statistics ranging from 31.94 to 297.89, indicating no weak instrument bias. These variants mapped to ion channel‐related genes such as CASQ2, SCN5A and SCN10A (Table  2 ).

Table 2.

Variants from GWAS and EWAS acting as genetic proxies for QRS duration used as causality estimates with chronic heart failure

SNP	Chr	Pos (hg19)	Effect allele	Other allele	Beta	Standard error	P value	EAF	F statistics	Gene name
Genome‐wide association study (GWAS)
rs41312411	3	38621237	G	C	0.14	0.02	7.93E‐15	0.15	60.53	‐
rs4131768	3	38695174	A	G	0.08	0.01	1.03E‐09	0.68	37.34	‐
rs10054375	5	153871832	C	T	−0.09	0.01	3.84E‐11	0.36	43.79	‐
rs3176323	6	36646849	C	T	0.11	0.01	6.11E‐16	0.30	65.60	‐
rs10282556	7	116130983	A	G	−0.07	0.01	1.35E‐08	0.54	32.31	‐
rs2237844	8	17490906	G	A	−0.15	0.03	1.91E‐08	0.06	31.64	‐
Exome‐wide association study (EWAS)
rs17391905	1	51546140	G	T	−1.06	0.15	3.37E‐13	0.02	52.98	exm‐rs17391905
rs4074536	1	116310967	C	T	−0.33	0.05	2.20E‐11	0.29	44.78	CASQ2
rs7562790	2	36673555	G	T	0.35	0.05	2.03E‐14	0.41	58.51	CRIM1
rs17020136	2	37248015	C	T	0.39	0.07	1.09E‐08	0.20	32.68	HEATR5B
rs17362588	2	179721046	A	G	0.53	0.08	2.61E‐11	0.09	44.45	CCDC141
rs11710077	3	38657899	T	A	−0.94	0.06	1.13E‐58	0.20	260.84	SCN5A
rs6795970	3	38766675	A	G	0.79	0.05	9.49E‐67	0.40	297.89	SCN10A
rs4687718	3	53282303	A	G	−0.39	0.07	1.59E‐08	0.13	31.94	TKT
rs6762208	3	185331165	A	C	−0.31	0.05	4.88E‐11	0.35	43.22	SENP2
rs13165478	5	153869040	A	G	−0.68	0.05	6.77E‐47	0.36	206.83	exm‐rs13165478
rs9470361	6	36623379	A	G	0.88	0.05	4.00E‐63	0.24	281.25	exm‐rs9470361
rs11153730	6	118667522	C	T	0.58	0.05	9.38E‐38	0.49	164.95	exm‐rs11153730
rs1362212	7	35305306	A	G	0.55	0.06	1.83E‐18	0.15	76.86	exm‐rs1362212
rs7784776	7	46620145	G	A	0.26	0.05	1.14E‐08	0.40	32.58	exm‐rs7784776
rs3807989	7	116186241	A	G	0.40	0.05	2.88E‐18	0.40	75.97	CAV1
rs16898691	8	124663987	G	C	−0.94	0.12	3.97E‐15	0.04	61.71	KLHL38
rs2958149	12	57109792	A	G	−0.32	0.05	3.61E‐10	0.27	39.31	NACA
rs3825214	12	114795443	G	A	0.48	0.06	1.81E‐17	0.20	72.34	TBX5
rs7966651	12	115371958	T	C	−0.42	0.05	9.30E‐17	0.27	69.11	exm‐rs7966651
rs1886512	13	74520186	A	T	−0.40	0.05	2.08E‐14	0.37	58.45	KLF12
rs11848785	14	72057355	G	A	−0.45	0.05	2.17E‐18	0.25	76.53	SIPA1L1
rs4966020	15	99284680	G	A	−0.28	0.05	2.63E‐09	0.36	35.44	IGF1R
rs17608766	17	45013271	C	T	0.71	0.07	8.68E‐27	0.14	114.81	GOSR2
rs9912468	17	64318357	G	C	0.42	0.05	7.51E‐19	0.42	78.63	PRKCA
rs663651	18	42456653	G	A	−0.46	0.05	2.88E‐17	0.44	71.43	SETBP1
rs961253	20	6404281	A	C	0.30	0.05	1.07E‐10	0.36	41.70	exm‐rs961253
rs3746429	20	33703607	T	C	−0.37	0.06	3.56E‐10	0.17	39.34	EDEM2

beta, the regression coefficient, also known as the effect size; Chr, chromosome; EAF, effect allele frequency; F statistics, A measure of the strength of the instrument in Mendelian randomization studies (for instrumental variables); Pos (hg19), position (on hg19); SNP, single nucleotide polymorphism; standard error, the standard error of the beta estimate.

3.3. A shorter QRS duration predicts a higher CHF risk

The causal inference was performed using the identified genetic instruments for QRS duration. IVW (MRE) results showed a genetically predicted shorter QRS duration associated with an increased risk of CHF [OR (95% CI) = 0.77 (0.62, 0.96), P = 0.02]. Weighted median [OR (95% CI) = 0.81 (0.70, 0.94), P = 0.01], IVW radial [OR (95% CI) = 0.77 (0.62, 0.97), P = 0.02] and RAPS [OR (95% CI) = 0.82 (0.68, 0.99), P = 0.04] further supported this finding. For EWAS, MR analysis with IVW, IVW radial and RAPS showed consistent causal effects with smaller error margins [OR (95% CI) = 0.98 (0.96, 0.99), P = 0.03] (Figure  1 ). Non‐zero intercept values in MR Egger suggested that QRS duration and CHF causal estimates were unlikely influenced by horizontal pleiotropy (Table  3 ).

Figure 1.

MR estimates using variants from GWAS and EWAS as genetic instruments of QRS duration with chronic heart failure. (A) Forest plots presenting causal results estimated using different MR models. (B) Scatter plot showing regression coefficients for individual variants in causal estimates. 95% CI, 95% confidence interval; EWAS, exome‐wide association study; GWAS, genome‐wide association study; IVW, inverse variance weighted; OR, odds ratio; SNP, single nucleotide polymorphism.

Table 3.

Horizontal pleiotropy evaluation in MR causal estimation of QRS duration and chronic heart failure using MR Egger regression

QRS data	Egger_intercept	Standard error	P value
GWAS	0.025	0.05	0.67
EWAS	0.015	0.01	0.23
Corrected GWAS	0.002	0.02	0.94
Corrected EWAS	0.007	0.01	0.38

Egger_intercept, the intercept term from the MR‐Egger regression model.

Leave‐one‐out analysis indicated no single SNP driving the overall effect in the GWAS group, while removing rs9470361 in the EWAS group showed significant effects, suggesting its potentially large influence (Figure 2 A ). Therefore, radial MR was used to identify and adjust for potential outliers in MR analysis. The radial plot identified two outliers (rs10054375 and rs3176323) for GWAS and 10 outliers, including rs9470361, for EWAS (Figure 2 B ). Re‐evaluation excluding outliers showed more significant MR estimates in both GWAS and EWAS [GWAS: IVW (OR (95% CI) = 0.84 (0.78, 0.91), P = 2.26E‐05), weighted median (OR (95% CI) = 0.81 (0.69, 0.94), P = 5.71E‐03) and EWAS: IVW (OR (95% CI) = 0.98 (0.96, 0.99), P = 6.57E‐05), weighted median (OR (95% CI) = 0.98 (0.96, 0.99), P = 0.01)] (Figure 3 A ). The corrected leave‐one‐out analysis for EWAS data showed no single SNP significantly driving the overall MR effect (Figure 3 B ). RAPS, handling overdispersion and pleiotropy, produced diagnostic plots indicating that the included genetic instruments conformed to the MR normal distribution assumption, with residuals meeting model assumptions (F value: 1.54, P value: 0.21), suggesting that the causal estimate between QRS duration and CHF is reliable (Figure 3 C ).

Figure 2.

Results of sensitivity analysis of the causal relationship between QRS duration and chronic heart failure. (A) Leave‐one‐out sensitivity analysis showing the effect of individual variants (GWAS, left panel; EWAS, right panel) variants on the estimates. (B) Pleiotropic outlier‐detected radial MR plot of QRS duration GWAS (left panel) and EWAS (right panel) variants on chronic heart failure. The x‐axis represents the square root of the weight for each variant. The y‐axis shows the radial estimate. Detected outliers are displayed in different colours.

Figure 3.

Outlier corrected MR analysis. (A) Forest plot showing results of deleted post hoc analysis of potential pleiotropic variants in the causal assessment of QRS duration and chronic heart failure. (B) Leave‐one‐out analysis after variance‐corrected analysis in EWAS. The red line represents the total risk, as indicated by the odds ratio. (C) Diagnostic plots of MR normal distribution test after variance corrected analysis in EWAS using MR‐RAPS.

4. Discussion

This study utilized GWAS and EWAS data to investigate the genetic determinants of QRS duration and its causal relationship with CHF. The main findings indicate that genetically determined shorter QRS duration is associated with an increased risk of CHF, suggesting a causal link between the two. These results provide new insights into the role of genetic components in QRS duration and their impact on CHF risk, potentially improving the clinical diagnosis and management of heart failure patients.

Our findings align with previous research indicating a significant genetic contribution to QRS duration. Heritability estimates from GWAS data showed a genetic variance of 16.3%, while EWAS data showed a slightly higher estimate of 18.5%. This suggests that coding regions might have a more direct functional impact on cardiac electrical activity. Prolonged QRS duration is observed frequently in patients with heart failure and is associated with poor prognosis and adverse outcomes in terms of left ventricular remodelling and survival, especially in patients not receiving CRT. ¹² , ¹³ , ¹⁴ , ¹⁵ Our study extends these findings by demonstrating that even shorter QRS duration can predict higher CHF risk. Randomized controlled trials (RCTs) such as the EchoCRT study have shown that CRT does not reduce mortality or heart failure hospitalization rates in CHF patients with QRS duration less than 130 ms and may even increase mortality. ¹⁹ , ²⁰ Therefore, CRT is contraindicated in patients with shorter QRS durations. ²¹ Our results follow these findings, suggesting that narrow QRS duration may become an important indicator of prognostic management in CHF.

The discrepancy between the detrimental effects of prolonged QRS duration and the predictive value of shorter QRS duration for CHF might be attributed to differences in study design, population characteristics, and the use of genetic tools for causal inference. The association between shorter QRS duration and increased CHF risk may be driven by underlying genetic structures affecting cardiac electrical conduction. Our study identified variants in ion channel‐related genes (e.g., CASQ2, SCN5A and SCN10A) that play critical roles in cardiac electrophysiology. These genetic determinants likely influence ventricular depolarization efficiency and ventricular remodelling, thereby affecting heart failure risk. ²² , ²³ , ²⁴ , ²⁵ For example, variants in SCN5A, encoding the Nav1.5 channel, have been associated with electrical remodelling of sodium channels in CHF cardiomyocytes, which also display previously unrecognized non‐electrical‐producing roles and may affect the structural integrity of the heart, leading to cardiomyopathy. ²³ , ²⁶ GWAS identified novel variants in the SCN5A/SCN10A locus identified as controlling myocardial mass and affecting heart failure risk. ²⁷ Zhang et al. reported that a synonymous SNP (rs1805126) in the SCN5A gene regulates its interaction with miR‐24, thereby suppressing both its own expression and the function of Nav1.5. ²⁶ Additionally, the extremely low heritability of CHF highlights the significant role of non‐genetic factors such as hypertension, diabetes and lifestyle choices in its development.

The use of MR provided a robust tool for inferring causality, with multiple MR methods and sensitivity analyses supporting the robustness of our causal inference. These methods minimize the risk of confounding and reverse causation inherent in observational studies, thereby strengthening the validity of our findings. However, this study has limitations. One potential limitation of this study is the reliance on predominantly European cohorts, which may limit the generalizability of our findings to other populations. Genetic diversity among different ethnic groups can influence the prevalence of certain variants, as well as their effects on phenotype. For instance, studies have shown that genetic variants in SCN5A, CASQ2 and SCN10A may have different frequencies or functional impacts across populations of African, Asian or Latin American descent. ²⁸ Therefore, while our findings are valuable for European populations, further studies involving diverse cohorts are needed to assess whether the observed associations between genetic variants and QRS duration, as well as the risk for CHF, hold true across different genetic backgrounds. Addressing this gap in research will be crucial for expanding the clinical applicability of these findings globally. Excluding patients with QRS duration greater than 120 ms and other comorbid conditions might also limit the applicability of our findings to a broader CHF population. Although MR methods aim to minimize pleiotropy, the possibility of residual pleiotropy (where variants affect multiple traits) cannot be completely ruled out. ¹⁸ Further research is needed to replicate these findings in diverse populations and explore the biological mechanisms underlying the observed associations.

In conclusion, this study establishes a causal relationship between shorter QRS duration and increased CHF risk. Identifying QRS duration as a genetic marker for CHF risk can enhance early diagnosis and personalized treatment strategies. These findings have significant implications for the clinical management of heart failure, emphasizing the necessity of a comprehensive assessment of QRS duration in patient management. Future research should focus on replicating these findings and elucidating the underlying mechanisms to explore therapeutic targets for informed interventions and improved patient outcomes.

5. Conclusions

This study provides robust evidence for a causal relationship between reduced QRS duration and elevated risk of CHF. Recognizing QRS duration as a genetic marker for CHF risk offers valuable insights for improving early diagnosis and guiding precision treatment strategies in clinical practice.

Funding

This work was supported by the National Natural Science Foundation of China (NSFC) (Grant Number 82300347).

Conflict of interest

None declared.

Consent for publication

All authors approved the final version and agreed to be responsible for the study.

Zheng, Z. , Li, X. , and Song, Y. (2025) Heritability and causality of QRS duration and chronic heart failure risk. ESC Heart Failure, 12: 3018–3027. 10.1002/ehf2.15321.

Data availability statement

The data underlying this article are available in the article and more detailed data can be attained by consulting the authors.