Skip to main content
  • Loading metrics

Multiple anthropometric measures and proarrhythmic 12-lead ECG indices: A mendelian randomization study

  • Maddalena Ardissino ,

    Contributed equally to this work with: Maddalena Ardissino, Kiran Haresh Kumar Patel

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Visualization, Writing – original draft

    ‡ These authors share first authorship on this work.

    Affiliations National Heart and Lung Institute, Imperial College London, London, United Kingdom, Royal Papworth Hospital, Cambridge Biomedical Campus, Cambridge, United Kingdom

  • Kiran Haresh Kumar Patel ,

    Contributed equally to this work with: Maddalena Ardissino, Kiran Haresh Kumar Patel

    Roles Data curation, Investigation, Methodology, Project administration, Validation, Visualization, Writing – original draft

    ‡ These authors share first authorship on this work.

    Affiliation National Heart and Lung Institute, Imperial College London, London, United Kingdom

  • Bilal Rayes,

    Roles Data curation, Formal analysis, Validation, Writing – review & editing

    Affiliation National Heart and Lung Institute, Imperial College London, London, United Kingdom

  • Rohin K. Reddy,

    Roles Conceptualization, Investigation, Methodology, Project administration, Visualization, Writing – original draft

    Affiliation National Heart and Lung Institute, Imperial College London, London, United Kingdom

  • Greg J. Mellor,

    Roles Conceptualization, Investigation, Methodology, Writing – review & editing

    Affiliation Royal Papworth Hospital, Cambridge Biomedical Campus, Cambridge, United Kingdom

  • Fu Siong Ng

    Roles Conceptualization, Investigation, Supervision, Visualization, Writing – review & editing

    Affiliation National Heart and Lung Institute, Imperial College London, London, United Kingdom



Observational studies suggest that electrocardiogram (ECG) indices might be influenced by obesity and other anthropometric measures, though it is difficult to infer causal relationships based on observational data due to risk of residual confounding. We utilized mendelian randomization (MR) to explore causal relevance of multiple anthropometric measures on P-wave duration (PWD), PR interval, QRS duration, and corrected QT interval (QTc).

Methods and findings

Uncorrelated (r2 < 0.001) genome-wide significant (p < 5 × 10−8) single nucleotide polymorphisms (SNPs) were extracted from genome-wide association studies (GWAS) on body mass index (BMI, n = 806,834), waist:hip ratio adjusted for BMI (aWHR, n = 697,734), height (n = 709,594), weight (n = 360,116), fat mass (n = 354,224), and fat-free mass (n = 354,808). Genetic association estimates for the outcomes were extracted from GWAS on PR interval and QRS duration (n = 180,574), PWD (n = 44,456), and QTc (n = 84,630). Data source GWAS studies were performed between 2018 and 2022 in predominantly European ancestry individuals. Inverse-variance weighted MR was used for primary analysis; weighted median MR and MR-Egger were used as sensitivity analyses. Higher genetically predicted BMI was associated with longer PWD (β 5.58; 95%CI [3.66,7.50]; p = < 0.001), as was higher fat mass (β 6.62; 95%CI [4.63,8.62]; p < 0.001), fat-free mass (β 9.16; 95%CI [6.85,11.47]; p < 0.001) height (β 4.23; 95%CI [3.16, 5.31]; p < 0.001), and weight (β 8.08; 95%CI [6.19,9.96]; p < 0.001). Finally, genetically predicted BMI was associated with longer QTc (β 3.53; 95%CI [2.63,4.43]; p < 0.001), driven by both fat mass (β 3.65; 95%CI [2.73,4.57]; p < 0.001) and fat-free mass (β 2.08; 95%CI [0.85,3.31]; p = 0.001). Additionally, genetically predicted height (β 0.98; 95%CI [0.46,1.50]; p < 0.001), weight (β 3.45; 95%CI [2.54,4.36]; p < 0.001), and aWHR (β 1.92; 95%CI [0.87,2.97]; p = < 0.001) were all associated with longer QTc. The key limitation is that due to insufficient power, we were not able to explore whether a single anthropometric measure is the primary driver of the associations observed.


The results of this study support a causal role of BMI on multiple ECG indices that have previously been associated with atrial and ventricular arrhythmic risk. Importantly, the results identify a role of both fat mass, fat-free mass, and height in this association.

Author summary

Why was the study done?

  • Previous studies have shown that a higher body mass index (BMI) is associated with changes in electrocardiogram (ECG) measurements.
  • However, it is unclear from available research whether this is driven by fat mass or by lean mass because BMI does not differentiate between these.
  • Additionally, most available data come from observational studies, which are liable to confounding and reverse causation. This limitation can be overcome by the use of mendelian randomization (MR)

What did the researchers do and find?

  • In this MR study, the authors explored the association between higher genetically predicted BMI, fat mass, fat free mass, weight, height, and waist:hip ratio and ECG measures of atrial and ventricular conduction.
  • The authors found that higher genetically predicted BMI was associated with many ECG indices that have been linked to arrhythmic risk and that this was driven by fat mass, fat free mass, and height.

What do the findings mean?

  • In this study, lean body mass traits appear to correlate with proarrhythmic electrophysiological remodeling.
  • These data may help to risk stratify individuals who would benefit from targeted weight management interventions.
  • The main limitations are inclusion of data from only predominantly European ancestry populations, and the inability to establish whether a single body size measure might be the most important in driving the associations.


Obesity is a growing global health problem associated with high rates of cardiometabolic morbidity [13] and is an independent risk factor for atrial [4] and ventricular arrhythmias [5]. Previous observational studies have identified an association between increasing adiposity and abnormalities on the 12-lead electrocardiogram (ECG) [6], such as longer P-wave duration (PWD) and increased P-wave dispersion (signifying delayed atrial conduction) [7,8] and increased QT dispersion and prolongation of corrected QT interval (QTc) (markers of abnormal ventricular activation and repolarization) [911]. These ECG changes are of clinical importance, as prior mendelian randomization (MR) studies have demonstrated their causal relevance on multiple adverse outcomes such as atrial fibrillation (AF) [12] and sudden cardiac death (SCD) [13].

However, body mass index (BMI) is not only determined by adiposity. Additional anthropometric measures, such as fat mass, fat-free mass, and height, are important determinants of BMI. Although it is thought that the association between BMI and electrophysiological remodeling is strongly mediated by adipose tissue, the other determinants of BMI above may also play an important role, and the potential influence of these parameters on ECG indices remains relatively less explored. Additionally, beyond overall adiposity, its distribution might play an important role in determining arrhythmic risk. This is clinically relevant given that visceral fat is thought to be more detrimental to health than subcutaneous adiposity [14]. From the current well-established association of BMI with proarrhythmic ECG changes, it is therefore impossible to establish whether it is truly only increasing adiposity that influences electrophysiological remodeling, and beyond this, it is difficult to establish whether this association is truly causal and independent of other concurrent cardiometabolic risk factors. For example, higher BMI is associated with multiple morbidities such as hypertension and ischemic heart disease, which also induce adverse cardiac remodeling [3]. In turn, these associations may also be at least partly mediated by an intermediate risk factor, such as type 2 diabetes. In the setting of observational data, it is difficult to accurately characterize these complex causal relationships.

MR is a genetic epidemiological method that can provide estimates of the effects of risk factors on an outcome using a framework that is less liable to influence by reverse causation and confounding [15]. MR leverages the random process of allele assortment and conception, which leads to an effective “randomization” of individuals to high or low genetic risk of a risk factor. This genetic liability can then be used as a proxy for the exposure itself, in an instrumental variable analysis framework. Since the initial genetic risk allocation is almost entirely random, similar to randomization in a clinical trial, it is not influenced by confounding or reverse causation. Thus, the MR framework can be used to investigate the causal relevance of the exposure on an outcome under a set of key assumptions.

The aim of this study was to utilize MR to explore the association of anthropometric metrics and on PWD, PR interval, QRS duration, and QTc interval, to elucidate the effect of body size and composition on proarrhythmic electrophysiological remodeling. Uniquely, in addition to the commonly used metric of BMI, we sought to investigate the specific effects of its determinants of height, fat mass, and fat-free mass and to explore the influence of adiposity distribution through exploring BMI-adjusted waist:hip ratio (aWHR). At present, despite the extensive observational evidence linking higher BMI to changes in ECG measures, no study, observational or MR based, has explored this.


Ethics and data access

This study used publicly available genome-wide association summary data. All included studies had gained ethical approval and participant consent according to individual protocols available at the referenced publications. The study is reported on the basis of the Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization (STROBE-MR) Guidelines (S1 STROBE Checklist) [16]. All analyses were carried out on R version 4.1.2 [17] using the TwoSampleMR [18] and Mendelianrandomization packages [19]. No protocol was preregistered.

Instrumental variable selection

Genome-wide significant (p < 5 × 10−8) instrumental variables for BMI (in kg/m2) [20], aWHR (in cm:cm) [20], and height (in inverse normal transformed and standardized cm) [21] were extracted from summary statistics of published genome-wide association studies (GWAS), respectively, including 806,834 participants, 697,734 participants, and 709,594 participants. Genome-wide significant (p < 5 × 10−8) instrumental variables for weight (in inverse normal rank transformed kg, n = 360,116), whole-body fat mass (in inverse normal rank transformed kg, n = 354,244), and whole-body fat-free mass (in inverse normal rank transformed kg, n = 354,808) were extracted from Neale lab UK Biobank GWAS summary statistics ( Whole-body fat mass and fat-free mass were measured using impedance, which were measured as outlined in the protocol and original paper [22].

Genetic association estimates for the outcomes were extracted from GWAS of outcomes, including PR interval and QRS duration (ms, n = 180,574 participants) [23], PWD (n = 44,456 participants) [24], and QTc (n = 84,630) [25]. The ECG measures in the original GWAS studies were derived from 12-lead ECG at rest taken in the supine position. Extraction of ECG measures, such as P-wave indices, was performed using a variety of software algorithms that are described in detail in the individual GWAS papers [2325].

All GWAS studies for both exposures and outcomes were performed within the scope of large, multicenter population-based studies. Further details on study cohorts are available at the respective publications and are provided in Table 1.

Table 1. Information on the studies and consortia from which genetic association data were obtained.

Harmonization and clumping

Gene-exposure association estimates for each exposure were harmonized with gene-outcome association estimates for corresponding instrumental SNPs in the outcome data. An attempt was made to infer positive strand alleles during harmonization. Where this could not be inferred, or if SNPs were palindromic or ambiguous, the SNP was excluded. Only SNPs with available gene-exposure and gene-outcome association estimates were included; if there were no matching SNPs for an instrumental variable in the outcome data, proxies were not sought. After harmonization, SNPs were clumped to retain only uncorrelated variants (pairwise linkage disequilibrium r2 < 0.001). Instrument strength was quantified using F-statistics.

Statistical analysis

Inverse-variance weighted (IVW) MR with multiplicative random effects [26] was used as the primary analysis method for all models, to estimate the association between each genetically predicted anthropometric trait and ECG phenotype [15]. Results are presented as beta coefficients (β) with respective 95% confidence intervals (95%CI), which can be interpreted as the expected change in ECG interval, in ms, per 1-unit change in the exposure. The units for each exposure are reported in Table 1.

There are 3 core assumptions of the IVW MR approach that, if not met, can lead to unreliable results. These include the following:

  1. That instrumental variables predict the exposure;
  2. That instrumental variables are not associated with confounders of the association between the exposure and outcome; and
  3. That instrumental variables are only associated with the outcome through the exposure.

The first assumption was tested by quantification of instrument strength using F-statistics. In instances where instrumental SNPs influence the outcome through additional biological pathways that are parallel to, but do not act through the exposure, these assumptions are violated in a phenomenon called horizontal pleiotropy. Sensitivity analysis using weighted median MR [27] and MR-Egger can be used to explore this phenomenon. The weighted median method has been shown to provide consistent estimates even in situations where up to half of the instrumental SNPs are invalid or horizontally pleiotropic [27]. Additionally, the MR-Egger method can be used to more formally test for the presence of directional pleiotropy through the addition and testing of an intercept term, though the method relies on the weaker assumption that the instrument strength is independent of direct effects (InSIDE assumption) [28].

Statistical significance was considered at an alpha value of 0.0021 after Bonferroni adjustment for testing of 24 hypotheses (6 exposures on 4 outcomes, 0.05/24).


P-wave duration

Higher genetically predicted BMI was associated with longer PWD (β 5.58; 95%CI [3.66,7.50]; p < 0.001), as was higher genetically predicted fat mass (β 6.62; 95%CI [4.63,8.62]; p < 0.001) and fat-free mass (β 9.16; 95%CI [6.85,11.47]; p < 0.001). Genetically predicted height (β 4.23; 95%CI [3.16, 5.31]; p < 0.001) and weight (β 8.08; 95%CI [6.19,9.96]; p < 0.001) were also associated with longer PWD. However, genetically predicted aWHR was not associated with differences in PWD (β −1.24; 95%CI [−3.73,1.25]; p = 0.330). The results are displayed in Fig 1 and Table 2.

Fig 1. MR estimates for the effects of genetically predicted BMI on PWD and PR interval.

aWHR, adjusted waist:hip ratio; BMI, body mass index; CI, confidence interval; MR, mendelian randomization; ms, milliseconds; PWD, P-wave duration.

Table 2. Univariable MR estimates for the effects of genetically predicted anthropometric traits on ECG parameters, using an IVW model with multiplicative random effects.

As displayed in Table 3, the results remained consistent on sensitivity analyses using weighted median MR. MR-Egger estimates also remained consistent, and intercept test did not identify evidence of directional pleiotropy in the association with BMI (p = 0.534), height (p = 0.107), fat mass (p = 0.657), fat-free mass (p = 0.326), weight (p = 0.292), and aWHR (p = 0.748). Instrumental variable F-statistics were >10 in all cases, as reported in Table 4.

Table 3. MR sensitivity analyses for effects of genetically predicted anthropometric traits on ECG parameters, using weighted median MR and MR-Egger models.

PR interval

Higher genetically predicted BMI was associated with longer PR interval (β 2.29; 95%CI [0.78,3.79]; p = 0.003), but this was no longer significant after accounting for multiple testing (threshold p-value = 0.0021). No association was observed between higher genetically predicted fat mass (β −0.97; 95%CI [−2.61,0.67]; p = 0.248), fat-free mass (β 0.66; 95%CI [−1.29,2.60]; p = 0.509), height (β −0.52; 95%CI [−1.41,0.38]; p = 0.258), weight (β 0.42; 95%CI [−1.11,1.94]; p = 0.594), or aWHR (β −0.24; 95%CI [−2.00,1.52]; p = 0.791) and PR interval. The results are displayed in Fig 1 and Table 2.

As displayed in Table 3, the results remained consistent on sensitivity analyses using weighted median MR and MR-Egger. MR-Egger intercept test did not identify evidence of directional pleiotropy in the association with BMI (p = 0.498), height (p = 0.451), fat mass (p = 0.631), fat-free mass (p = 0.913), weight (p = 0.689), and aWHR (p = 0.078). F-statistics for all instruments were >10, as reported in Table 4.

QRS duration

Higher genetically predicted BMI was not associated differences in QRS interval (β 0.06; 95%CI [−0.62,0.74]; p = 0.856) and neither were genetically predicted fat mass (β −0.20; 95%CI [−0.96,0.55]; p = 0.599), fat-free mass (β −0.01; 95%CI [−1.35,1.33]; p = 0.989), height (β −0.59; 95%CI [−1.52,0.34]; p = 0.211), weight (β −0.24; 95%CI [−1.08,0.59]; p = 0.567), and aWHR (β 0.69; 95%CI [−1.09,2.48]; p = 0.445). The results are displayed in Fig 2 and Table 2.

Fig 2. MR estimates for the effects of genetically predicted BMI on QRS duration and QT interval.

aWHR, adjusted waist:hip ratio; BMI, body mass index; CI, confidence interval; MR, mendelian randomization; ms, milliseconds.

As displayed in Table 3, the results remained consistent on sensitivity analyses using weighted median MR. MR-Egger estimates also remained consistent, and intercept test did not identify evidence of directional pleiotropy in the association with BMI (p = 0.802), height (p = 0.263), fat mass (p = 0.537), fat-free mass (p = 0.927), weight (p = 0.987), and aWHR (p = 0.076). F-statistics for all instruments were >10, as reported in Table 4.

Corrected QT interval

Higher genetically predicted BMI was associated with longer QTc (β 3.53; 95%CI [2.63,4.43]; p < 0.001), driven by both genetically predicted fat mass (β 3.65; 95%CI [2.73,4.57]; p < 0.001) and fat-free mass (β 2.08; 95%CI [0.85,3.31]; p = 0.001). Additionally, genetically predicted height (β 0.98; 95%CI [0.46,1.50]; p < 0.001), weight (β 3.45; 95%CI [2.54,4.36]; p < 0.001), and aWHR (β 1.92; 95%CI [0.87,2.97]; p < 0.001) were all associated with longer QTc. The results are displayed in Fig 2 and Table 2.

As displayed in Table 3, the results remained consistent on sensitivity analyses using weighted median MR. MR-Egger estimates also remained consistent, and intercept test did not identify evidence of directional pleiotropy in the association with BMI (p = 0.709), height (p = 0.088), fat mass (p = 0.704); fat-free mass (p = 0.939), weight (p = 0.305), and aWHR (p = 0.575). F-statistics for all instruments were >10, as reported in Table 4.


In this study, we used MR to investigate the causal relevance of multiple anthropometric traits, relating to adiposity and lean body composition, and 12-lead ECG indices that are associated with atrial and ventricular arrhythmias. The main findings may be summarized in 4 key points. First, the results support a causal nature of the association between adiposity and prolonged PWD but do not support a causal role of adiposity on PR interval. Second, they reveal an association between higher lean mass and height with PWD but not PR interval. Third, higher adiposity was associated with a prolonged QTc interval but was not associated with a longer QRS interval. Finally, we demonstrate an association between lean mass and height with prolonged QTc, which has not been previously reported.

The causal inferences that can be made on the basis of MR study results is reliant on meeting a number of instrumental variant assumptions. These state that the instruments must predict the exposure, that they must be associated with the outcome only through the exposure, and that no confounders or common causes of the exposure, in this case the genetic variants, and the outcome exist. In order to address the first assumption, we formally quantified the F-statistics and detected no evidence of weak instruments that might bias the results. To evaluate the second assumption, sensitivity analyses were carried out. These only identified evidence of horizontal pleiotropy for the association of BMI with PR interval, suggesting that this association might be biased, and, therefore, this result should not be taken to support causal relevance. Importantly, however, it is known that MR-Egger performs less well in settings where there is significant overlap between the GWAS studies used to source gene-exposure and gene-outcome association estimates. Since this is the case in the present study, it cannot be excluded that some pleiotropy might exist for other associations [29]. To further investigate, we also performed weighted median MR analyses to corroborate findings from MR-Egger, and this provides further reassurance that directional pleiotropy is unlikely to have occurred undetected. The final assumption cannot formally be tested but can be mitigated by the use of GWAS data from similar ancestries. For this reason, we only utilized data from studies in European ancestry individuals. However, for some data sources, specifically that of PR interval and QRS duration, there was a proportion of non-European ancestry individuals included in the study due to lack of availability of data in European ancestry individuals only. Though only a minority of participants were non-European ancestry in these studies, it should be considered that some bias might arise from this.

The results of this study support a causal nature of the association between BMI and slower atrial conduction. Prior observational studies have outlined an association between adiposity traits, including weight, BMI, and fat mass, on atrial depolarization, slower atrial conduction, most commonly describing longer PWD [8,11,3033] in individuals with higher BMI. The results of our study suggests that weight control is likely to play an important role in reduction of atrial arrhythmic risk in individuals with obesity and optimization of rhythm control strategies in AF. Promising evidence has established a degree of reversibility of these atrial ECG changes following weight loss [8,3437]. Combined with prior observational evidence linking longer PWD with greater AF risk and higher recurrence rates after catheter ablation [3846], our results provide further evidence to support the central role of weight reduction to reduce both the risk of onset and recurrence of atrial arrhythmias.

Importantly, the results suggest that adiposity traits predominantly associate with PWD, with no evidence association with PR interval. We therefore suggest that the potential influence of anthropometric traits is likely predominantly though electrophysiological and structural remodeling of the atria (e.g., fibrosis, dilatation) [47], resulting in slower atrial conduction, rather than through direct influence on the specialized conduction tissue, as the AV node is the major determinant of PR interval. Mechanistically, epicardial fat might play a key role, as it is known to influence cardiac electrophysiology by modulating ionic currents via paracrine mechanisms [6]. Importantly, as previously noted, the GWAS for PR interval was in a multi-ancestry cohort, so the null findings should not be interpreted as conclusive evidence of lack of association and should be replicated in a European ancestry cohort when this becomes available.

In contrast to the results of our study, a prior MR investigation identified an inverse, rather than direct, association between PWD, PR interval, and AF risk [12]. This is not intuitively in agreement with our results. Higher BMI has a well-established causal effect on AF. In this study, we also demonstrate that higher BMI causes slower atrial depolarization, with higher PWD and PR interval. Thus, we would expect that higher, rather than lower, PWD and PR interval might be associated with higher risk of AF. This discrepancy is most likely explained by the existence of a nonlinear association between atrial ECG indices and AF risk, whereby both individuals at the low and high end of the distribution are at increased risk of AF. However, given the current lack of evidence to demonstrate this, it is an important target for further research as duly highlighted by the authors in the original study.

Together with the findings on adiposity-related traits, our results support an additional role of both fat-free mass and height on increased PWD. Height and fat-free body mass are 2 closely correlated phenotypes [48]. Although the association between these anthropometric traits and ECG indices of atrial conduction has not been described previously, studies have reported a stronger contribution of lean body mass on excess AF risk conferred by BMI [4952] compared to fat mass. Similarly, height has been associated with greater risk of AF in observational and MR studies [5361]. Although the mechanisms underlying these associations are unclear, a possible explanation relates to left atrial volume. Indeed, the association between left atrial volumes and AF risk is well established [62]. Taller individuals have higher absolute left atrial volumes since cardiac chamber volume is a direct function of body size [63]. It is possible that, despite the nonpathological nature of the increased volume conferred by height, it nevertheless contributes to a degree of electrophysiological dysfunction and arrhythmia predisposition [63], as suggested by the association with longer PWD described in the MR analyses this study. Overall, the mechanism behind this association remains an important unanswered question and a key target for further exploration.

There were no associations between any of the anthropomorphic traits studied and QRS duration. Since ventricular depolarization is facilitated by the His–Purkinje network, this suggests that the effects of body composition are more pronounced on ventricular myocardium than on specialized conduction tissue. The observational evidence of associations between adiposity and QRS duration is mixed, with some studies finding no association [64] and another reporting an association between increasing BMI and QRS duration that was independent of other covariates such as sex and age [65]. Our study adds to this contrasting evidence by suggesting a lack of causal association between adiposity and QRS duration. However, it must be noted that the GWAS study for QRS interval was performed a multi-ancestry cohort; therefore, the null findings should not be interpreted as conclusive evidence of lack of association and should be replicated in a European ancestry cohort when this becomes available.

Several observational studies have described direct associations between BMI and QTc [9,11,66], with a degree of reversibility after significant weight loss [66]. Another recent observational study using UK Biobank data also showed that QTc interval prolongs with increasing BMI, body fat, waist:hip ratio, as well as hip and waist circumference [67]. The association between longer QTc and risk of SCD is well established and supported by MR evidence [13]. In line with this, obesity has been associated with a higher risk of ventricular arrhythmias in prior observational studies [68] and is the most common ischemic cause of SCD [69] even after adjusting for age, sex, ethnicity, and cardiovascular risk factors [14]. The results of the present study support that this previously described association might be of causal relevance. The results specifically demonstrate an association between BMI and QTc, and additionally of both total fat mass and aWHR with QTc, suggesting an important role of both the volume of adiposity and its distribution. Considering the established association between QTc and SCD, as well as the reversibility in ECG phenotypes demonstrated after significant weight loss, the results herein stress the importance of weight reduction on reducing risk of SCD in individuals with obesity. Though SCD remains a rare cause of mortality, targeted weight loss intervention in those at high-risk, such as those with prior myocardial infarction or reduced ejection fraction, is likely of important benefit and should be a key priority for randomized study.

The association of lean body mass and height with ventricular repolarization is less well described than that with adiposity. To date, there are no published studies, observational or MR based, assessing the association of height and lean body mass with ventricular ECG parameters. However, observational evidence exists describing of greater burden of ventricular ectopics with increasing height [70,71]. Conversely, one prior study has described an inverse association between height and SCD [72]. Pathophysiologically, lean body mass and height, which are closely correlated, are known to be major predictors of absolute left ventricular mass [73]. Left ventricular mass and its diagnostic correlate of left ventricular hypertrophy are, in turn, known to be associated with longer QTc interval [7477] and are predictors of SCD, including among young, healthy individuals and athletes [7881]. Mechanistically, the associations between height and fat-free mass with longer QTc observed in this study may be mediated by what would be described as a “physiological” increase in left ventricular mass. Overall, the results raise the possibility that taller individuals with a greater lean body mass have a more proarrhythmic electroanatomic ventricular architecture.

There are a number of limitations to this study. First, the effect estimates presented cannot be used to infer the predicted change in ECG parameters per unit increase or decrement in the exposure phenotypes with the exceptions of BMI and aWHR. This is because the inverse normal rank transformation carried out on the exposure measures for analysis in the original GWAS studies mean that the unit changes in the exposure cannot be intuitively interpreted. Second, the outcome data sources for PR interval and QRS duration were from a GWAS on a population of mixed ancestry, while exposure data sets were on populations of European ancestry. Though both exposure and outcome estimates were adjusted for population structure, there can be residual bias in the MR estimates from population stratification. This can act in either direction (both away from the null or toward the null), and, therefore, the null results on the analyses on these outcomes should not be taken as unequivocal evidence of lack of an association. Third, we attempted to further explore the mechanistic role of left atrial size and ventricular mass in the associations demonstrated in this study, but there were insufficient instruments to carry out MR analysis in the largest available GWAS to date [82]. Fourth, we attempted to perform multivariable analyses including multiple anthropometric traits that are known to be correlated (e.g., fat mass and fat-free mass) in an attempt to establish the key driving factors in the associations with PWD and QTc. Unfortunately, the majority of these analyses had weak instruments (conditional F-statistics <10) and therefore were unreliable. The results of this analysis are reported in S1 Table This remains an important target for future research once larger GWAS data becomes available. Fifth, the use of predominantly European ancestry specific data limits generalisability to other ancestries; the analysis should therefore be replicated using data from other ancestries once available. Sixth, in this study, there was a degree of sample overlap, as both the exposure and outcome data sets included UK Biobank participants. The potential bias that might stem from this is, however, limited, as sample overlap has been shown to exert a very limited influence on results in the setting of large biobanks even when complete sample overlap exists [29]. Finally, the GWAS studies that were used to derive gene-outcome association data utilized a range of methods and software programs to extract the ECG indices. Within the scope of our study, we did not have access to individual-level data and therefore could not access the ECG tracings, and for this reason, we were unable to recalculate the ECG indices using different techniques such as signal averaging, which is known to be particularly useful in individuals with obesity.

Our results support causal associations between multiple anthropometric traits, both adiposity- and non-adiposity related, and abnormal ECG indices that have been previously causally related to risk of atrial and ventricular arrhythmias. The results stress the mechanistic importance of weight management in at-risk individuals to prevent or reverse proarrhythmic electrophysiological remodeling. The majority of associations related to ECG indices reflect electrophysiological function of cardiomyocytes rather than function of specialized conducting tissue. The results additionally identify an important role of fat-free mass and height, which have not been previously described as proarrhythmic factors except in the setting of AF risk. These are key targets for further investigation, as they might provide important biological insight and aid clinical risk stratification in high-risk individuals.

Supporting information

S1 STROBE Checklist. Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization (STROBE-MR) checklist of recommended items to address in mendelian randomization (MR) studies.


S1 Table. Results of multivariable mendelian randomization (MVMR) analyses.



The authors would like to acknowledge the participants and investigators of the GIANT, UK Biobank, ARIC, HCHS/SOL, MESA, and WHI studies.


  1. 1. Tsur AM, Twig G. The actual burden of obesity—accounting for multimorbidity. Lancet Diabetes Endocrinol. 2022;10:233–234. pmid:35248170
  2. 2. Dai H, Alsalhe TA, Chalghaf N, Riccò M, Bragazzi NL, Wu J. The global burden of disease attributable to high body mass index in 195 countries and territories, 1990–2017: An analysis of the Global Burden of Disease Study. Wareham NJ, editor. PLoS Med. 2020;17:e1003198. pmid:32722671
  3. 3. Larsson SC, Bäck M, Rees JMB, Mason AM, Burgess S. Body mass index and body composition in relation to 14 cardiovascular conditions in UK Biobank: a Mendelian randomization study. Eur Heart J. 2020;41:221–226. pmid:31195408
  4. 4. Lavie CJ, Pandey A, Lau DH, Alpert MA, Sanders P. Obesity and Atrial Fibrillation Prevalence, Pathogenesis, and Prognosis. J Am Coll Cardiol. 2017;70:2022–2035.
  5. 5. Sabbag A, Sidi Y, Kivity S, Beinart R, Glikson M, Segev S, et al. Obesity and exercise-induced ectopic ventricular arrhythmias in apparently healthy middle aged adults. Eur J Prev Cardiol. 2016;23:511–517. pmid:26069245
  6. 6. Patel KHK, Hwang T, Se Liebers C, Ng FS. Epicardial adipose tissue as a mediator of cardiac arrhythmias. Am J Physiol Circ Physiol. 2022;322:H129–H144. pmid:34890279
  7. 7. Yılmaz M, Altın C, Tekin A, Erol T, Arer İ, Nursal TZ, et al. Assessment of Atrial Fibrillation and Ventricular Arrhythmia Risk after Bariatric Surgery by P Wave/QT Interval Dispersion. Obes Surg. 2018;28:932–938. pmid:28900850
  8. 8. Russo V, Ammendola E, De Crescenzo I, Docimo L, Santangelo L, Calabrò R. Severe Obesity and P-Wave Dispersion: The Effect of Surgically Induced Weight Loss. Obes Surg. 2008;18:90–96. pmid:18080825
  9. 9. Omran J, Bostick BP, Chan AK, Alpert MA. Obesity and Ventricular Repolarization: a Comprehensive Review. Prog Cardiovasc Dis. 2018;61:124–135. pmid:29698642
  10. 10. Kumar T, Jha K, Sharan A, Sakshi P, Kumar S, Kumari A. Study of the effect of obesity on QT-interval among adults. J Fam Med Prim Care. 2019;8:1626. pmid:31198727
  11. 11. Seyfeli E, Duru M, Kuvandık G, Kaya H, Yalcin F. Effect of obesity on P-wave dispersion and QT dispersion in women. Int J Obes (Lond). 2006;30:957–961. pmid:16432544
  12. 12. Gajendragadkar PR, von Ende A, Ibrahim M, Valdes-Marquez E, Camm CF, Murgia F, et al. Assessment of the causal relevance of ECG parameters for risk of atrial fibrillation: A mendelian randomisation study. PLoS Med. 2021;18:1–18. pmid:33983917
  13. 13. Young WJ, Lahrouchi N, Isaacs A, Duong T, Foco L, Ahmed F, et al. Genetic analyses of the electrocardiographic QT interval and its components identify additional loci and pathways. Nat Commun. 2022;13:5144. pmid:36050321
  14. 14. Adabag S, Huxley RR, Lopez FL, Chen LY, Sotoodehnia N, Siscovick D, et al. Obesity related risk of sudden cardiac death in the atherosclerosis risk in communities study. Heart. 2015;101:215–221. pmid:25410499
  15. 15. Burgess S, Davey Smith G, Davies NM, Dudbridge F, Gill D, Glymour MM, et al. Guidelines for performing Mendelian randomization investigations. Wellcome Open Res. 2020;4:186. pmid:32760811
  16. 16. Davey Smith G, Davies N, Dimou N, Egger M, Gallo V, Golub R, et al. STROBE-MR: Guidelines for strengthening the reporting of Mendelian randomization studies. 2019.
  17. 17. R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2021.
  18. 18. Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D, et al. The MR-Base platform supports systematic causal inference across the human phenome. Elife. 2018:7. pmid:29846171
  19. 19. Yavorska OO, Burgess S. MendelianRandomization: an R package for performing Mendelian randomization analyses using summarized data. Int J Epidemiol. 2017;46:1734–1739. pmid:28398548
  20. 20. Pulit SL, Stoneman C, Morris AP, Wood AR, Glastonbury CA, Tyrrell J, et al. Meta-analysis of genome-wide association studies for body fat distribution in 694 649 individuals of European ancestry. Hum Mol Genet. 2019;28:166–174. pmid:30239722
  21. 21. Yengo L, Sidorenko J, Kemper KE, Zheng Z, Wood AR, Weedon MN, et al. Meta-analysis of genome-wide association studies for height and body mass index in ∼700000 individuals of European ancestry. Hum Mol Genet. 2018;27:3641–3649. pmid:30124842
  22. 22. Franssen FME, Rutten EPA, Groenen MTJ, Vanfleteren LE, Wouters EFM, Spruit MA. New Reference Values for Body Composition by Bioelectrical Impedance Analysis in the General Population: Results From the UK Biobank. J Am Med Dir Assoc. 2014;15:448.e1–448.e6. pmid:24755478
  23. 23. Wojcik GL, Graff M, Nishimura KK, Tao R, Haessler J, Gignoux CR, et al. Genetic analyses of diverse populations improves discovery for complex traits. Nature. 2019;570:514–518. pmid:31217584
  24. 24. Christophersen IE, Magnani JW, Yin X, Barnard J, Weng L-C, Arking DE, et al. Fifteen Genetic Loci Associated With the Electrocardiographic P Wave. Circ Cardiovasc Genet. 2017:10. pmid:28794112
  25. 25. Nauffal V, Morrill VN, Jurgens SJ, Choi SH, Hall AW, Weng L-C, et al. Monogenic and Polygenic Contributions to QTc Prolongation in the Population. Circulation. 2022;145:1524–1533. pmid:35389749
  26. 26. Burgess S, Bowden J, Fall T, Ingelsson E, Thompson SG. Sensitivity analyses for robust causal inference from mendelian randomization analyses with multiple genetic variants. Epidemiology. 2017;28:30–42. pmid:27749700
  27. 27. Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent Estimation in Mendelian Randomization with Some Invalid Instruments Using a Weighted Median Estimator. Genet Epidemiol. 2016;40:304–314. pmid:27061298
  28. 28. Burgess S, Thompson SG. Interpreting findings from Mendelian randomization using the MR-Egger method. Eur J Epidemiol. 2017;32:377–389. pmid:28527048
  29. 29. Minelli C, Del Greco M F, van der Plaat DA, Bowden J, Sheehan NA, Thompson J. The use of two-sample methods for Mendelian randomization analyses on single large datasets. Int J Epidemiol. 2021;50:1651–1659. pmid:33899104
  30. 30. Cosgun M, Sincer I, Inanir M, Erdal E, Mansiroglu AK, Gunes Y. P-wave Duration and Dispersion in Lone Obesity. J Coll Physicians Surg Pak. 2021;30:567–570. pmid:34027870
  31. 31. Magnani JW, Lopez FL, Soliman EZ, Maclehose RF, Crow RS, Alonso A. P wave indices, obesity, and the metabolic syndrome: the atherosclerosis risk in communities study. Obesity (Silver Spring). 2012;20:666–672. pmid:21475136
  32. 32. Liu T, Fu Z, Korantzopoulos P, Zhang X, Wang S, Li G. Effect of obesity on p-wave parameters in a Chinese population. Ann Noninvasive Electrocardiol. 2010;15:259–263. pmid:20645969
  33. 33. Faramawi MF, Delhey L, Abouelenein S, Delongchamp R. Metabolic syndrome and P-wave duration in the American population. Ann Epidemiol. 2020;46:5–11. pmid:32532371
  34. 34. Ibisoglu E, Boyraz B, Güneş ST, Savur Ü, Naki Tekin DD, Erdoğan A, et al. Impact of surgical weight loss on novel P-wave-related variables which are nominated as predictors of atrial arrhythmias. Pacing Clin Electrophysiol. 2021;44:1516–1522. pmid:34312874
  35. 35. Falchi AG, Grecchi I, Muggia C, Tinelli C. Weight loss and P wave dispersion: a preliminary study. Obes Res Clin Pract. 2014:e614–e617. pmid:25240700
  36. 36. Duru M, Seyfeli E, Kuvandik G, Kaya H, Yalcin F. Effect of weight loss on P wave dispersion in obese subjects. Obesity (Silver Spring). 2006;14:1378–1382. pmid:16988080
  37. 37. Vedel-Larsen E, Iepsen EW, Lundgren J, Graff C, Struijk JJ, Hansen T, et al. Major rapid weight loss induces changes in cardiac repolarization. J Electrocardiol. 2016;49:467–472. pmid:26925492
  38. 38. Nielsen JB, Pietersen A, Graff C, Lind B, Struijk JJ, Olesen MS, et al. Risk of atrial fibrillation as a function of the electrocardiographic PR interval: Results from the Copenhagen ECG Study. Heart Rhythm. 2013;10:1249–1256. pmid:23608590
  39. 39. Park J, Kim T, Lee JS, Park JK, Uhm JS, Joung B, et al. Prolonged PR Interval Predicts Clinical Recurrence of Atrial Fibrillation After Catheter Ablation. J Am Heart Assoc. 2014:3. pmid:25292186
  40. 40. Hari KJ, Nguyen TP, Soliman EZ. Relationship between P-wave duration and the risk of atrial fibrillation. Expert Rev Cardiovasc Ther. 2018;16:837–843. pmid:30295096
  41. 41. Schumacher K, Dagres N, Hindricks G, Husser D, Bollmann A, Kornej J. Characteristics of PR interval as predictor for atrial fibrillation: association with biomarkers and outcomes. Clin Res Cardiol. 2017;106:767–775. pmid:28382425
  42. 42. Demirtas AO, Icen YK, Donmez Y, Koca H, Kaypakli O, Koc M. Silent atrial fibrillation is associated with P-wave duration index in patients with cardiac resynchronisation therapy. Arch Med Sci Atheroscler Dis. 2019;4:e74–e81. pmid:31211273
  43. 43. Smith JW, O’Neal WT, Shoemaker MB, Chen LY, Alonso A, Whalen SP, et al. PR-Interval Components and Atrial Fibrillation Risk (from the Atherosclerosis Risk in Communities Study). Am J Cardiol. 2017;119:466–472. pmid:27889043
  44. 44. Kaypakli O, Koca H, Şahin DY, Okar S, Karataş F, Koç M. Association of P wave duration index with atrial fibrillation recurrence after cryoballoon catheter ablation. J Electrocardiol. 2018;51:182–187. pmid:29146378
  45. 45. Li R, Yang X, Jia M, Wang D, Cui X, Bai L, et al. Effectiveness of P-wave ECG index and left atrial appendage volume in predicting atrial fibrillation recurrence after first radiofrequency catheter ablation. BMC Cardiovasc Disord. 2021;21:164. pmid:33823799
  46. 46. Hong M, Hwang I, Yu H-T, Kim T-H, Uhm J-S, Joung B, et al. Potential causal association of a prolonged PR interval and clinical recurrence of atrial fibrillation after catheter ablation: a Mendelian randomization analysis. J Hum Genet. 2020;65:813–821. pmid:32409696
  47. 47. Alpert MA, Karthikeyan K, Abdullah O, Ghadban R. Obesity and Cardiac Remodeling in Adults: Mechanisms and Clinical Implications. Prog Cardiovasc Dis. 2018;61:114–123. pmid:29990533
  48. 48. Burton RF. Relationships among fat mass, fat-free mass and height in adults: A new method of statistical analysis applied to NHANES data. Am J Hum Biol Off J Hum Biol Counc. 2017:29. pmid:27862528
  49. 49. Fenger-Grøn M, Overvad K, Tjønneland A, Frost L. Lean Body Mass Is the Predominant Anthropometric Risk Factor for Atrial Fibrillation. J Am Coll Cardiol. 2017;69:2488–2497. pmid:28521886
  50. 50. Frost L, Benjamin EJ, Fenger-Grøn M, Pedersen A, Tjønneland A, Overvad K. Body fat, body fat distribution, lean body mass and atrial fibrillation and flutter. A Danish cohort study. Obesity (Silver Spring). 2014;22:1546–1552. pmid:24436019
  51. 51. Azarbal F, Stefanick ML, Assimes TL, Manson JE, Bea JW, Li W, et al. Lean body mass and risk of incident atrial fibrillation in post-menopausal women. Eur Heart J. 2016;37:1606–1613. pmid:26371115
  52. 52. Worm MS, Bager CL, Blair JPM, Secher NH, Riis BJ, Christiansen C, et al. Atrial fibrillation is associated with lean body mass in postmenopausal women. Sci Rep. 2020;10:573. pmid:31953421
  53. 53. Levin MG, Judy R, Gill D, Vujkovic M, Verma SS, Bradford Y, et al. Genetics of height and risk of atrial fibrillation: A Mendelian randomization study. Rienstra M, editor. PLoS Med. 2020;17:e1003288. pmid:33031386
  54. 54. Marott JL, Skielboe AK, Dixen U, Friberg JB, Schnohr P, Jensen GB. Increasing population height and risk of incident atrial fibrillation: the Copenhagen City Heart Study. Eur Heart J. 2018;39:4012–4019. pmid:29961878
  55. 55. Zia I, Johnson L, Memarian E, Borné Y, Engström G. Anthropometric measures and the risk of developing atrial fibrillation: a Swedish Cohort Study. BMC Cardiovasc Disord. 2021;21:602. pmid:34922449
  56. 56. Kerola T, Dewland TA, Vittinghoff E, Heckbert SR, Stein PK, Marcus GM. Predictors of atrial ectopy and their relationship to atrial fibrillation risk. Europace. 2019;21:864–870. pmid:30843034
  57. 57. Park YM, Moon J, Hwang IC, Lim H, Cho B. Height is associated with incident atrial fibrillation in a large Asian cohort. Int J Cardiol. 2020;304:82–84. pmid:31954587
  58. 58. Lai FY, Nath M, Hamby SE, Thompson JR, Nelson CP, Samani NJ. Adult height and risk of 50 diseases: a combined epidemiological and genetic analysis. BMC Med. 2018;16:187. pmid:30355295
  59. 59. Mont L, Tamborero D, Elosua R, Molina I, Coll-Vinent B, Sitges M, et al. Physical activity, height, and left atrial size are independent risk factors for lone atrial fibrillation in middle-aged healthy individuals. Europace. 2008;10:15–20. pmid:18178694
  60. 60. Andersen K, Rasmussen F, Neovius M, Tynelius P, Sundström J. Body size and risk of atrial fibrillation: a cohort study of 1.1 million young men. J Intern Med. 2018;283:346–355. pmid:29178512
  61. 61. Rosenberg MA, Kaplan RC, Siscovick DS, Psaty BM, Heckbert SR, Newton-Cheh C, et al. Genetic variants related to height and risk of atrial fibrillation: the cardiovascular health study. Am J Epidemiol. 2014;180:215–222. pmid:24944287
  62. 62. Knecht S, Pradella M, Reichlin T, Mühl A, Bossard M, Stieltjes B, et al. Left atrial anatomy, atrial fibrillation burden, and P-wave duration-relationships and predictors for single-procedure success after pulmonary vein isolation. Europace. 2018;20:271–278. pmid:28339545
  63. 63. Olsen FJ, Møgelvang R, Modin D, Schnohr P, Jensen GB, Biering-Sørensen T. Association between Isometric and Allometric Height-Indexed Left Atrial Size and Atrial Fibrillation. J Am Soc Echocardiogr. 2022;35:141–150.e4. pmid:34757164
  64. 64. Chi P-C, Chang S-C, Yun C-H, Kuo J-Y, Hung C-L, Hou CJ-Y, et al. The Associations between Various Ectopic Visceral Adiposity and Body Surface Electrocardiographic Alterations: Potential Differences between Local and Remote Systemic Effects. Pizzi C, editor. PLoS ONE. 2016;11:e0158300. pmid:27391045
  65. 65. Rao ACA, Ng ACC, Sy RW, Chia KKM, Hansen PS, Chiha J, et al. Electrocardiographic QRS duration is influenced by body mass index and sex. IJC Heart Vasc. 2021;37:100884. pmid:34660881
  66. 66. Omran J, Firwana B, Koerber S, Bostick B, Alpert MA. Effect of obesity and weight loss on ventricular repolarization: A systematic review and meta-analysis. Obes Rev. 2016;17:520–530. pmid:26956255
  67. 67. Patel KHK, Li X, Xu X, Sun L, Ardissino M, Punjabi PP, et al. Increasing Adiposity Is Associated With QTc Interval Prolongation and Increased Ventricular Arrhythmic Risk in the Context of Metabolic Dysfunction: Results From the UK Biobank. Front Cardiovasc Med. 2022;9:939156. pmid:35845082
  68. 68. Pietrasik G, Goldenberg I, McNitt S, Moss AJ, Zareba W. Obesity As a Risk Factor for Sustained Ventricular Tachyarrhythmias in MADIT II Patients. J Cardiovasc Electrophysiol. 2007;18:181–184. pmid:17338766
  69. 69. Aune D, Schlesinger S, Norat T, Riboli E. Body mass index, abdominal fatness, and the risk of sudden cardiac death: a systematic review and dose–response meta-analysis of prospective studies. Eur J Epidemiol. 2018;33:711–722. pmid:29417316
  70. 70. Ahmed S, Hisamatsu T, Kadota A, Fujiyoshi A, Segawa H, Torii S, et al. Ventricular Premature Complexes and Their Associated Factors in a General Population of Japanese Men. Am J Cardiol. 2022;169:51–56. pmid:35045928
  71. 71. von Rotz M, Aeschbacher S, Bossard M, Schoen T, Blum S, Schneider S, et al. Risk factors for premature ventricular contractions in young and healthy adults. Heart. 2017;103:702–707. pmid:27798051
  72. 72. Rosenberg MA, Lopez FL, Bůžková P, Adabag S, Chen LY, Sotoodehnia N, et al. Height and risk of sudden cardiac death: the Atherosclerosis Risk in Communities and Cardiovascular Health Studies. Ann Epidemiol. 2014;24:174–179.e2. pmid:24360853
  73. 73. Bella JN, Devereux RB, Roman MJ, O’Grady MJ, Welty TK, Lee ET, et al. Relations of Left Ventricular Mass to Fat-Free and Adipose Body Mass. Circulation. 1998;98:2538–2544.
  74. 74. Oikarinen L, Nieminen MS, Toivonen L, Viitasalo M, Wachtell K, Papademetriou V, et al. Relation of QT interval and QT dispersion to regression of echocardiographic and electrocardiographic left ventricular hypertrophy in hypertensive patients: the Losartan Intervention For Endpoint Reduction (LIFE) study. Am Heart J. 2003;145:919–925. pmid:12766755
  75. 75. Porthan K, Virolainen J, Hiltunen TP, Viitasalo M, Väänänen H, Dabek J, et al. Relationship of electrocardiographic repolarization measures to echocardiographic left ventricular mass in men with hypertension. J Hypertens. 2007;25:1951–1957. pmid:17762661
  76. 76. Mukerji R, Terry BE, Fresen JL, Petruc M, Govindarajan G, Alpert MA. Relation of left ventricular mass to QTc in normotensive severely obese patients. Obesity (Silver Spring). 2012;20:1950–1954. pmid:21818155
  77. 77. Salles GF, Cardoso CRL, Deccache W. Multivariate associates of QT interval parameters in diabetic patients with arterial hypertension: importance of left ventricular mass and geometric patterns. J Hum Hypertens. 2003;17:561–567. pmid:12874614
  78. 78. Laukkanen JA, Khan H, Kurl S, Willeit P, Karppi J, Ronkainen K, et al. Left ventricular mass and the risk of sudden cardiac death: a population-based study. J Am Heart Assoc. 2014;3:e001285. pmid:25376188
  79. 79. Han H-C, Parsons SA, Teh AW, Sanders P, Neil C, Leong T, et al. Characteristic Histopathological Findings and Cardiac Arrest Rhythm in Isolated Mitral Valve Prolapse and Sudden Cardiac Death. J Am Heart Assoc. 2020;9:e015587. pmid:32233752
  80. 80. Konety SH, Koene RJ, Norby FL, Wilsdon T, Alonso A, Siscovick D, et al. Echocardiographic Predictors of Sudden Cardiac Death: The Atherosclerosis Risk in Communities Study and Cardiovascular Health Study. Circ Cardiovasc Imaging. 2016:9. pmid:27496550
  81. 81. Haider AW, Larson MG, Benjamin EJ, Levy D. Increased left ventricular mass and hypertrophy are associated with increased risk for sudden death. J Am Coll Cardiol. 1998;32:1454–1459. pmid:9809962
  82. 82. Aung N, Vargas JD, Yang C, Cabrera CP, Warren HR, Fung K, et al. Genome-wide analysis of left ventricular image-derived phenotypes identifies fourteen loci associated with cardiac morphogenesis and heart failure development. Circulation. 2019;140:1318–1330. pmid:31554410