Submitted:
19 July 2026
Posted:
22 July 2026
You are already at the latest version
Abstract
Background: Glucagon-like peptide-1 (GLP-1) and dual GLP-1/glucose-dependent insulinotropic polypeptide (GIP) receptor agonists reduce cardiovascular events, but the underlying cardiac structural remodeling remains unclear. Objectives: This study evaluated longitudinal cardiac remodeling associated with semaglutide and tirzepatide in a real-world cohort and determined its weight-loss dependency. Methods: Using electronic health records from a federated network, we analyzed longitudinal echocardiograms of patients prescribed semaglutide or tirzepatide, stratified by 12-month weight loss into super-responders (>15%), moderate-responders (5%-15%), and minimal-responders (<5%). Additionally, GLP-1/GIP patients with >5% weight loss were propensity-matched with non-GLP-1/GIP weight-loss medication control patients on demographics and baseline body mass index. Results: Among 3,500 GLP-1/GIP patients (422 weight-loss super-responders, 1,426 moderate-responder, 1,652 minimal-responders), left ventricular (LV) mass decreased across all groups proportional to weight loss (super-responders: 198.4 ± 69.6 to 176.6 ± 66.0 g; moderate-responders: 210.7 ± 74.0 to 195.8 ± 68.1 g; minimal-responders: 216.1 ± 68.4 to 204.8 ± 67.4 g; p < 0.001). Left atrial volume and LV systolic and diastolic function did not change. Among weight-loss super-responders, semaglutide (n = 159) showed greater LV mass reduction than tirzepatide (n = 89) (Cohen’s d = -0.389 vs. -0.263), while tirzepatide showed small improvements in right ventricular function. Compared to matched non-GLP-1/GIP controls (n = 118/group), GLP-1/GIP patients exhibited significant LV mass reduction not observed in controls despite similar weight loss. Conclusions: Real-world findings demonstrate that incretin-based therapies are associated with reverse cardiac remodeling characterized by weight-loss-dependent and incretin-specific mechanisms, with potentially distinct structural targets among agents.
Keywords:
incretin
; GLP-1/GIP
; cardiometabolic disease
; cardiac remodeling
; echocardiography
Introduction
Obesity is an important risk factor for the development of cardiovascular disease [1]. Recent clinical trials have demonstrated that glucagon-like peptide-1 (GLP-1) receptor agonists and dual GLP-1/glucose-dependent insulinotropic polypeptide (GIP) receptor agonists, such as semaglutide and tirzepatide, reduce the risk of adverse cardiovascular events in patients with cardiometabolic syndromes [2,3,4]. The Semaglutide Effects on Cardiovascular Outcomes in People with Overweight or Obesity (SELECT) trial showed that semaglutide lowers cardiovascular mortality in patients with obesity or preexisting cardiovascular disease [2]. In patients with heart failure with preserved ejection fraction (HFpEF), the Semaglutide Treatment Effect in People With Obesity and HFpEF (STEP-HFpEF) and Study of Tirzepatide in Participants with HFpEF and Obesity (SUMMIT) trials demonstrated reductions in cardiovascular mortality and heart failure (HF) exacerbations and improvements in HF symptoms and functional capacity in patients with HfpEF [3,4].
Despite these clinical observations, the cardiac structural remodeling that underpins these benefits remains incompletely understood. Preclinical studies have suggested that incretin-based therapies attenuate adverse cardiac remodeling via multiple pathways [5]. In animal models of HFpEF, GLP-1 and dual GLP-1/GIP receptor agonists have been shown to decrease left ventricular (LV) hypertrophy, myocardial fibrosis, and markers of systemic inflammation, while improving diastolic function [6,7,8]. Furthermore, these cardiac structural changes with incretin-based therapies exceeded the benefits achieved through dietary caloric restriction alone, suggesting that weight-loss-independent pathways may also play a role. In patients with obesity-related HFpEF, the echocardiography substudy of STEP-HFpEF demonstrated that semaglutide therapy improved left atrial (LA) volume, LV diastolic function, and RV size [9]. The cardiac magnetic resonance (CMR) substudy of the SUMMIT trial showed a reduction in LV mass and paracardiac adipose tissue with tirzepatide therapy [10].
However, key clinical questions persist: (1) how do these clinical trial findings translate to diverse, real-world populations receiving GLP-1 and dual GLP-1/GIP receptor agonists; (2) to what extent is cardiac structural remodeling linked to systemic weight loss or incretin-specific mechanisms; and (3) are there differential cardiac remodeling profiles associated with GLP-1 versus dual GLP-1/GIP receptor agonists? To address these knowledge gaps, we utilized longitudinal, de-identified data from a federated electronic health record (EHR) network to evaluate multi-chamber cardiac remodeling using longitudinal echocardiography in a real-world population of patients on semaglutide or tirzepatide therapy. We stratified patients by their 12-month weight-loss response to delineate whether cardiac structural remodeling occurs in a weight-loss-dependent or -independent fashion. Furthermore, we evaluated these changes against a propensity-matched control cohort of patients prescribed non-incretin anti-obesity medications to gain insights into the specific cardiac structural impacts of GLP-1/GIP receptor pathways relative to other weight-loss medications.
Methods
Data Source and Study Design
Study Cohorts
Two parallel, active-comparator treatment cohorts were constructed: an exposure cohort of patients prescribed semaglutide or tirzepatide, and an active control cohort of patients prescribed non-incretin weight-loss medications.
For the exposure cohort, eligible patients were adults aged 18 years or older who initiated semaglutide or tirzepatide within the study period, with the index date defined as the date of the first documented medication order. Inclusion required at least one weight measurement within the 90 days preceding the index date and at least one measurement within 12 months following the index date, as well as a baseline echocardiogram within 2 years before the index date and a follow-up echocardiogram performed between 6 months and 4 years post-index date. Patients were excluded if they had a documented history of bariatric surgery, including gastric bypass, sleeve gastrectomy, gastric banding, or duodenal switch.
The control cohort was defined using identical inclusion and exclusion criteria, except for the index medication. The control index date was defined as the first prescription order for benzphetamine, bupropion/naltrexone, diethylpropion, lorcaserin, orlistat, phendimetrazine, phentermine, phentermine/topiramate, or setmelanotide. Patients were also excluded if they received a prescription or administration of any GLP-1 receptor agonist (exenatide, lixisenatide, liraglutide, dulaglutide, semaglutide) or dual GIP/GLP-1 receptor agonist (tirzepatide) at any time during the study period
Demographics and Clinical Covariates
Demographic covariates, including age, sex, race, and ethnicity, as well as baseline clinical characteristics, were assessed within the 12 months preceding the index date. Clinical characteristics were extracted using International Classification of Diseases 9 and 10 codes and natural language processing algorithms on unstructured data.
Echocardiography Assessment
Structured echocardiography measurements from the nference federated network were used to evaluate multi-chamber cardiac structural and functional remodeling. LA remodeling was assessed using LA volume. LV structural remodeling was evaluated by LV end-diastolic diameter (LVEDD), LV end-systolic diameter (LVESD), interventricular septal thickness, posterior wall thickness, and relative wall thickness. LV systolic function was assessed by LV ejection fraction (LVEF), and diastolic function was evaluated by medial and septal early diastolic mitral annular velocities (e') and their corresponding E/e' ratios. Right ventricular (RV) function and pulmonary hemodynamics were assessed by tricuspid annular plane systolic excursion, tricuspid annular systolic velocity, tricuspid regurgitation maximum velocity (TR Vmax), and estimated right ventricular systolic pressure (RVSP). Due to expected fluctuations in weight and consequently body surface area, non-indexed echocardiographic values were used where applicable.
Weight Response Classification
To evaluate cardiac structural remodeling relative to the magnitude of weight loss, patients in the GLP-1 and dual GLP-1/GIP receptor agonist cohort were stratified into three distinct response categories based on their maximum percentage weight change within 12 months post-indexing. Weight-loss super-responders were defined as weight loss greater than 15%, weight-loss moderate-responders as weight loss between 5% and 15%, and weight-loss minimal-responders as a weight loss of less than 5%. Baseline weight was defined as the measurement closest to the index date within the 90-day pre-treatment window, and the percentage weight change was calculated relative to this baseline for all subsequent measurements.
Propensity Matching
To evaluate whether GLP-1/GIP receptor pathways are associated with distinct cardiac structural remodeling, patients with greater than 5% weight loss in the GLP-1 and dual GLP-1/GIP receptor agonist cohorts were propensity score-matched to patients in the non-incretin weight-loss medication cohort based on age, sex, and baseline BMI. One-to-one nearest-neighbor matching without replacement was executed on the logit of the propensity score using a strict caliper width of 0.01 standard deviations of the logit.
Statistical Analysis
Baseline demographic and clinical characteristics across the three GLP-1 and dual GLP-1/GIP receptor agonist cohort weight-loss response groups were compared using a one-way analysis of variance for continuous variables, reported as mean ± standard deviation, and the Pearson chi-square test for categorical variables, reported as frequencies and percentages. Within-cohort longitudinal changes in echocardiographic parameters from baseline to follow-up were evaluated using paired t-tests. Standardized effect sizes were quantified using Cohen’s d, with an absolute value greater than 0.20 prespecified as the minimum threshold for clinical meaningfulness. All statistical tests were two-sided, and a p-value less than 0.05 was considered statistically significant. Statistical analyses were performed using Python Version 3.10.20.
De-Identification and HIPAA Compliance Certification
All patient data underwent a rigorous expert determination de-identification process before analysis to satisfy Health Insurance Portability and Accountability Act Privacy Rule requirements [45 CFR §164.514(b)(1)]. This framework employed a multilayered transformation approach for both structured data (cryptographic hashing of identifiers, date shifting, and geographic truncation) and unstructured clinical text (ensemble deep learning and rule-based methods with >99% recall for personally identifiable information detection) [13,14]. nference established secure data environments within each participating center, housing these de-identified patient data governed by expert determination. These de-identified data environments were specifically designed to enable data access and analysis without requiring Institutional Review Board oversight, approval, or exemption confirmation. Accordingly, informed consent and Institutional Review Board review were not required for this study.
Data Availability
This study involves the analysis of de-identified EHR data via the nference Federated EHR Network. Data shown and reported in this manuscript were extracted from this environment using an established protocol for data extraction, aimed at preserving patient privacy. The data has been de-identified pursuant to an expert determination in accordance with the Health Insurance Portability and Accountability Act Privacy Rule. Any data beyond what is reported in the manuscript, including but not limited to the raw EHR data, cannot be shared or released due to the parameters of the expert determination to maintain the data de-identification. The corresponding author should be contacted for additional details regarding the nference federated network.
Code Availability
The code used to conduct the analyses in this study cannot be made publicly available because it contains proprietary analytical workflows and was executed within a secure, access-controlled environment for analysis of de-identified electronic health record data. The analytic methods are described in the manuscript to enable interpretation and reproducibility of the study findings.
Results
Baseline Characteristics
A total of 3,500 patients prescribed semaglutide or tirzepatide, with both a baseline echocardiogram at 255 ± 201 days pre-therapy initiation and a follow-up echocardiogram at 676 ± 339 days post-therapy initiation (Figure S1), were included in the analysis. Stratification by maximum percentage weight change within 12 months post-indexing yielded 422 patients in the weight-loss super-responder group (19.67 ± 8.1 kg), 1,426 patients in the weight-loss moderate-responder group (9.84 ± 6.5 kg), and 1,652 patients in the weight-loss minimal-responder group (4.1 ± 6.6 kg). Weight-loss super-responders were significantly younger (61.0 ± 12.0 years) than weight-loss moderate and minimal responders (63.9 ± 11.9 years; p < 0.001) and were more frequently female (63.5%, 44.6%, and 41.5%, respectively; p < 0.001). A similar percentage of patients in all three groups had a baseline diagnosis of HF (weight-loss super-responder: 40.0%, weight-loss moderate-responder: 44.2%, and weight-loss minimal-responder: 42.6%; p = 0.285). The baseline prevalence of diabetes increased progressively across the weight-loss super-responder, weight-loss moderate-responder, and weight-loss minimal-responder cohorts (52.4%, 68.2%, and 74.8%, respectively; p < 0.001). Similar increasing trends across these respective cohorts were observed for coronary artery disease (34.4%, 45.0%, and 45.8%, respectively; p < 0.001) and chronic kidney disease (38.4%, 49.6%, and 46.5%, respectively; p < 0.001) (Table 1).
Longitudinal Echocardiographic Evaluation of Cardiac Remodeling in GLP-1/GLP Receptor Agonist
LA volume did not show clinically meaningful changes from baseline to follow-up in any of the three weight-loss response groups. Measurements remained stable among weight-loss super-responders (68.2 ± 26.8 to 66.3 ± 30.1 mL; p = 0.14, d = -0.09) and weight-loss minimal-responders (73.1 ± 27.7 to 72.5 ± 29 mL; p = 0.43, d = -0.024). Although weight-loss moderate-responders achieved statistical significance (from 73.6 ± 31.9 to 71 ± 31.5 mL; p = 0.001), the effect size was negligible (d = -0.110) (Table 2).
In contrast to LA volume, significant LV structural remodeling was observed. LV mass decreased significantly across all three cohorts: weight-loss super-responders (198.4 ± 69.6 to 176.6 ± 66.0 g; p < 0.001, d = -0.420), weight-loss moderate-responders (210.7 ± 74.0 to 195.8 ± 68.1 g; p < 0.001, d = -0.274), and weight-loss minimal-responders (216.1 ± 68.4 to 204.8 ± 67.4 g; p < 0.001, d = -0.213) (Table 2).
Consistent with LV mass, interventricular septal thickness decreased across all three groups, with effect sizes proportional to the degree of weight loss achieved. The greatest reduction was observed in weight-loss super-responders (11.1 ± 2.2 to 10.4 ± 2.3 mm; p<0.001, d = -0.312), followed by weight-loss moderate responders (11.1 ± 2.2 to 10.9 ± 2.2 mm; p < 0.001, d = -0.148) and weight-loss minimal responders (11.3 ± 2.1 to 11.1 ± 2.1 mm; p = 0.002, d = -0.089) (Table 2). A similar pattern was seen for LV posterior wall thickness, which declined in all cohorts (weight-loss super-responders: 10.6 ± 1.9 to 9.9 ± 1.9 mm; weight-loss moderate-responders: 10.7 ± 1.9 to 10.4 ± 1.9 mm; weight-loss minimal-responders: 10.9 ± 1.8 to 10.7 ± 1.9 mm; all p < 0.001) (Table 2). The effect size for LV posterior wall thickness was substantially larger in weight-loss super-responders (d = -0.329) compared to weight-loss moderate-responders (d = -0.147) and weight-loss minimal-responders (d = -0.138), further supporting a weight-loss-dependent pattern of LV hypertrophy regression.
LVEDD was reduced in weight-loss super-responders (48.7 ± 6.8 to 47.6 ± 6.6 mm; p < 0.001, d = -0.204) and weight-loss moderate-responders (50.5 ± 7.4 to 49.2 ± 7.1 mm; p < 0.001, d = -0.224), both surpassing the prespecified clinical relevance threshold of absolute d greater than 0.2. In contrast, the reduction in weight-loss minimal responders, although statistically significant, did not reach a meaningful effect size (d = -0.062). Systolic and diastolic functional parameters, including LV ejection fraction and septal and lateral E/e’ ratios, did not change significantly in any of the groups (Table 2).
In addition to left-sided changes, right-sided hemodynamics showed significant improvement that correlated with the degree of weight loss. TR Vmax decreased across all three groups, changing from 2.5 ± 0.4 to 2.4 ± 0.4 m/s in weight-loss super-responders (p < 0.001) and from 2.6 ± 0.4 to 2.5 ± 0.4 m/s in both weight-loss moderate-responders (p = 0.002) and weight-loss minimal-responders (p = 0.007). However, a clinically meaningful effect size was only observed in the weight-loss super-responder group (d = -0.268), indicating that right-sided pressure reduction was most substantial in patients with the greatest weight loss (Table 2). RVSP declined in weight-loss super-responders (33.2 ± 12.6 to 30.4 ± 10.0 mmHg; p < 0.001, d = -0.235) and weight-loss moderate-responders (34.4 ± 11.3 to 33.3 ± 11.7 mmHg; p = 0.018, d = -0.087) but did not reach statistical significance in weight-loss minimal-responders (p = 0.066), with a clinically meaningful effect size observed only among weight-loss super-responders.
Comparison of Longitudinal Echocardiographic Changes Between Semaglutide and Tirzepatide Super Responders
Among weight-loss super responders, we further examined whether the echocardiographic improvements identified in the overall cohort were driven by a specific agent by comparing semaglutide (n = 159) and tirzepatide (n = 89) patients. Baseline demographics and clinical characteristics, except for chronic kidney disease, were similar between both groups. Patients on semaglutide had a lower baseline BMI than those on tirzepatide (32.3 ± 3.7 vs. 33.9 ± 4.3, p = 0.004) (Table S1).
Semaglutide was associated with significant regression of LV wall thickness, with meaningful reductions in both interventricular septal thickness (10.9 ± 2.2 to 10.3 ± 2.1 mm; p = 0.001, d = -0.298) and LV posterior wall thickness (10.4 ± 1.8 to 9.8 ± 1.8 mm; p = 0.001, d = -0.299), whereas tirzepatide did not demonstrate significant changes in either parameter (Table 3). LV mass regression was also more pronounced with semaglutide (196.0 ± 67.9 to 177.6 ± 64.8 g; p < 0.001, d = -0.386) compared to tirzepatide (197.3 ± 53.4 to 184.2 ± 59.7 g; p = 0.032, d = -0.263).
Interestingly, both semaglutide and tirzepatide demonstrated reduction in LVEDD: semaglutide (49.0 ± 6.4 to 47.9 ± 6.3 mm; p = 0.008, d = -0.228) and tirzepatide (50.1 ± 5.5 to 48.7 ± 6.1 mm; p = 0.015, d = -0.284), with both agents meeting the combined significance and effect size threshold. (Table 3). Tirzepatide also produced a significant and clinically meaningful reduction in TR Vmax (2.5 ± 0.5 to 2.4 ± 0.4 m/s; p = 0.044, d = -0.286), whereas semaglutide did not meet the effect size threshold for this measure (d = -0.161, p = 0.131). These findings suggest potentially distinct but complementary patterns of cardiac benefit between these two agents.
Evaluation of Longitudinal Echocardiographic Changes Between GLP-1/GIP Receptor Agonists and Non-Incretin Weight-Loss Medications
To evaluate whether GLP-1/GIP receptor pathways are associated with cardiac structural remodeling that is distinct from non-incretin weight-loss medications, we compared propensity-matched cohorts of GLP-1/GIP receptor agonist patients and patients receiving non-incretin weight-loss medications (n = 118 for both), both restricted to those achieving ≥5% weight loss and matched on age, sex, and baseline BMI. Patients on GLP-1/GIP receptor agonists had more diabetes, chronic kidney disease, heart failure, and atrial fibrillation at baseline (Table S2).
Consistent with previous findings, in the semaglutide and tirzepatide cohort, four parameters demonstrated significant and clinically meaningful improvement: interventricular septal thickness (10.6 ± 2.0 to 10.2 ± 1.8 mm; p = 0.030, d = -0.229), LV posterior wall thickness (10.4 ± 1.7 to 9.6 ± 1.4 mm; p < 0.001, d = -0.443), LV relative wall thickness (42.8 ± 9.3 to 40.9 ± 7.5 mm; p = 0.046, d = -0.209), and LV mass (194.2 ± 58.3 to 171.7 ± 56.8 g; p < 0.001, d = -0.499) (Table 4). None of these parameters showed significant or meaningful changes in the non-incretin weight-loss medication cohort despite similar weight loss, suggesting a drug-specific effect on LV structural remodeling beyond that attributable to weight reduction alone.
In both cohorts, LVEDD and LVESD decreased: LVEDD (semaglutide and tirzepatide cohort: 49.6 ± 7.0 to 48.2 ± 6.8 mm, p = 0.006, d = -0.264; non-incretin weight-loss medications: 49.2 ± 4.9 to 47.0 ± 5.7 mm, p = 0.002, d = -0.402) and LVESD (semaglutide and tirzepatide cohort: 33.7 ± 7.0 to 32.6 ± 6.9 mm, p = 0.021, d = -0.229; non-incretin weight-loss medication cohort: 32.3 ± 5.3 to 31.0 ± 5.8 mm, p = 0.050, d = -0.254) (Table 4), indicating that these parameters may be driven by weight loss itself rather than a GLP-1/GIP receptor agonist-specific mechanism.
Discussion
Our retrospective study in a large, real-world cohort shows that incretin-based therapies are associated with favorable LV and right-sided cardiac remodeling. Our findings suggest that (1) LV mass regression is proportional to the magnitude of total body weight reduction; (2) there may be distinct differences in cardiac structural remodeling between semaglutide and tirzepatide; and (3) independent of systemic weight loss, there may be direct cardioprotective effects mediated via GLP-1/GIP receptor agonism (Central Illustration).
The weight-loss-dependent changes in LV mass and wall thickness observed in our study are consistent with the findings from the CMR substudy of the SUMMIT trial [10]. In the SUMMIT CMR substudy, tirzepatide therapy resulted in a significant reduction in LV mass that correlated with the magnitude of body weight loss in patients with obesity-related HFpEF. While that clinical trial evaluated these cardiac parameters at 52 weeks post-treatment, our real-world data demonstrate that these structural changes may be observed over a longer follow-up period (mean 676 days) in a real-world population. These changes were most pronounced in our weight-loss super-responder group (greater than 15% weight loss), suggesting that the degree of myocardial remodeling is at least partly associated with the magnitude of achieved weight loss. Interestingly, in our analysis of patients on a GLP-1/GIP receptor agonist who were propensity-matched with patients on non-incretin weight-loss medications with similar weight reduction, we did observe significant differences in LV mass attenuation, favoring the incretin cohort. This observation suggests that while weight loss is an important component of the effects of GLP-1/GIP receptor agonism, there may be direct drug actions on myocardial tissue that are weight-loss-independent. This is further supported by the fact that the GLP-1 receptor is expressed on the human heart [15].
In contrast to the LV structural remodeling observed in our study, we did not observe improvements in LA volume or LV diastolic function in our overall cohort, as was seen in the echocardiographic substudy of the STEP-HFpEF trials [9] or in preclinical studies [7]. In the STEP-HFpEF substudy, semaglutide therapy was associated with a significant attenuation of LA remodeling and improvements in parameters of LV diastolic function, without affecting LA or LV mechanics or LV remodeling. However, in our real-world cohort, LA volume did not show clinically meaningful changes from baseline to follow-up in any of the three weight-loss response groups, and mitral inflow-to-annular velocity (E/e') ratios remained unchanged across all groups. Additionally, in both our sub-analysis of patients on semaglutide and the overall cohort analysis of patients on either semaglutide or tirzepatide, there was an improvement in LV hypertrophy, which was not seen in the STEP-HFpEF substudy. These discrepancies may reflect the heterogeneity of real-world populations compared to clinical trial cohorts.
Notably, our head-to-head subgroup analysis comparing semaglutide and tirzepatide among weight-loss super-responders revealed differential associations in cardiac remodeling between the two agents. Semaglutide therapy was primarily associated with regression of LV wall thickness, whereas tirzepatide demonstrated a small, but significant reduction in LVEF, LVEDD, and RV function. These remodeling patterns potentially suggest that single and dual incretin agents may engage different pathways. These variations may stem from differences in downstream signaling cascades when combining GIP co-activation with GLP-1 pathways versus isolated GLP-1 receptor activation [5].
In conclusion, our real-world findings demonstrate that incretin-based therapies drive significant reverse cardiac remodeling characterized by an interplay of weight-loss-dependent and incretin-specific mechanisms, with potentially distinct structural targets among specific agents.
Study Limitations
Our study has several limitations inherent to its retrospective, observational design, which should be considered in the context of randomized clinical trial evidence on GLP-1 and dual GLP-1/GIP receptor agonists in cardiometabolic disease. Confounding, misclassification, and selection bias are intrinsic to nonrandomized study designs.
Using EHR data from a federated network carries the risk of data heterogeneity, unmeasured residual confounding, and missing clinical variables. Because medication data were derived from EHR orders and administrations, key treatment-related variables, including exact dosing, long-term medication adherence, treatment persistence, and temporal changes in therapy, could not be definitively verified. Since echocardiographic parameters were evaluated through routine clinical care rather than a standardized core laboratory, inter-operator variability across the network could introduce measurement noise that may have obscured subtle changes in echocardiographic parameters.
Furthermore, although propensity score matching improved balance in our secondary analysis of patients on GLP-1 and dual GLP-1/GIP receptor agonist and those on non-GLP-1 and dual GLP-1/GIP receptor agonist weight-loss medications, this approach resulted in the exclusion of a significant proportion of treated individuals.
Finally, our head-to-head comparison between semaglutide and tirzepatide and our matched active-control comparisons were limited by a relatively small sample size, and the absence of statistically significant associations for certain echocardiographic parameters may reflect limited statistical power.
Clinical Perspectives
Competency in Medical Knowledge
In a large, real-world cohort of patients treated with incretin-based therapies, semaglutide and tirzepatide were associated with significant reverse cardiac remodeling over an extended follow-up period. Reduction in LV hypertrophy and LV mass tracked proportionally with the magnitude of systemic weight loss, achieving the greatest structural impact in weight-loss super-responders. Head-to-head subgroup comparisons revealed that semaglutide and tirzepatide exhibit potentially differential effects on cardiac remodeling. Furthermore, propensity-matched analysis in patients on incretin therapy compared to those on non-incretin weight-loss medications with equivalent weight loss showed that improvements in concentric LV remodeling and LV mass were only seen on incretin therapies, suggesting these agents may exert direct cardioprotective effects that are independent of systemic weight loss.
Translational Outlook
Future investigations should elucidate the exact cellular and molecular signaling pathways that differentiate single versus dual incretin receptor activation on multi-chamber cardiac structure and function. Translational studies should focus on distinguishing whether distinct variations in cardiac remodeling phenotypes observed between semaglutide and tirzepatide stem from selective GLP-1 receptor activation versus combined GLP-1/GIP receptor co-activation. Finally, prospective, head-to-head clinical imaging trials are required to validate these phenotype-specific adaptations and guide personalized therapeutic selection in patients with cardiometabolic disease.
Supplementary Materials
The following supporting information can be downloaded at the website of this paper posted on Preprints.org.
Funding
The authors declare no funding.
Acknowledgments
The authors thank the nference engineering team for the development and operation of the nference federated network.
Conflicts of Interest Statement
Dr. Rajendran is a co-founder at NeuCures and receives consulting fees from nference. Drs. Saha, Krishnappa, Venkatakrishnan, Murugadoss, and Soundararajan are employees of nference, Inc., which conducts research collaborations with various biopharmaceutical companies, including Eli Lilly and Company and Novo Nordisk A/S, whose GLP-1/GIP receptor agonist products (tirzepatide and semaglutide) are included in this study. None of these companies, nor any other nference collaborator, funded, supported, or had any role in the independent study design, data acquisition, analysis, interpretation, manuscript preparation, or the decision to submit this work for publication. All analyses were conducted by the authors using de-identified electronic health record data. The authors declare no additional competing interests.
Abbreviations
| CMR | cardiac magnetic resonance |
| EF | ejection fraction |
| EHR | electronic health record |
| GLP-1 | glucagon-like peptide-1 |
| GIP | glucose-dependent insulinotropic polypeptide |
| HF | heart failure |
| HFpEF | heart failure with preserved ejection fraction |
| IRB | institutional review board |
| LA | left atrium |
| LV | left ventricle |
| LVEDD | left ventricular end-diastolic diameter |
| LVESD | left ventricular end-systolic diameter |
| TR | tricuspid regurgitation |
| RV | right ventricle |
| RVSP | right ventricular systolic pressure |
References
- Powell-Wiley TM, Poirier P, Burke LE, et al. Obesity and Cardiovascular Disease: A Scientific Statement From the American Heart Association. Circulation. 2021;143:e984–e1010. [CrossRef]
- Lincoff AM, Brown-Frandsen K, Colhoun HM, et al. Semaglutide and Cardiovascular Outcomes in Obesity without Diabetes. New England Journal of Medicine. 2023;389:2221–2232. [CrossRef]
- Kosiborod MN, Abildstrøm SZ, Borlaug BA, et al. Semaglutide in Patients with Heart Failure with Preserved Ejection Fraction and Obesity. New England Journal of Medicine. 2023;389:1069–1084. [CrossRef]
- Packer M, Zile MR, Kramer CM, et al. Tirzepatide for Heart Failure with Preserved Ejection Fraction and Obesity. New England Journal of Medicine. 2025;392:427–437. [CrossRef]
- Thomas J, Dagan M, Wang B, Gutman S, Kaye DM. Mechanisms of GLP-1 Receptor Agonists in HFpEF: Exploring Weight-Dependent and Independent Drivers of Therapeutic Benefit. Circulation: Heart Failure. 2026;19:e013279. [CrossRef]
- Hegedűs ZI, Jakab ME, Gergely TG, et al. Tirzepatide, a dual GIP/GLP1-receptor co-agonist preserves cardiac function and improves survival in angiotensin II-induced heart failure model in mice: comparison to liraglutide. Cardiovasc Diabetol. 2025;24:253. [CrossRef]
- Nguyen TD, Shingu Y, Amorim PA, Schenkl C, Schwarzer M, Doenst T. GLP-1 Improves Diastolic Function and Survival in Heart Failure with Preserved Ejection Fraction. J of Cardiovasc Trans Res. 2018;11:259–267. [CrossRef]
- Withaar C, Meems LMG, Nollet EE, et al. The Cardioprotective Effects of Semaglutide Exceed Those of Dietary Weight Loss in Mice With HFpEF. JACC: Basic to Translational Science. 2023;8:1298–1314. [CrossRef]
- Solomon SD, Ostrominski JW, Wang X, et al. Effect of Semaglutide on Cardiac Structure and Function in Patients With Obesity-Related Heart Failure. J Am Coll Cardiol. 2024;84:1587–1602. [CrossRef]
- Kramer CM, Borlaug BA, Zile MR, et al. Tirzepatide Reduces LV Mass and Paracardiac Adipose Tissue in Obesity-Related Heart Failure. JACC. 2025;85:699–706. [CrossRef]
- Lenehan P, Venkatakrishnan AJ, Murugadoss K, et al. Clinical nSights: A software platform to accelerate real world oncology analyses. J Clin Oncol. 2024;42:e23316–e23316. [CrossRef]
- Murugadoss K, Venkatakrishnan AJ, Gregg CJ, Soundararajan V. Semaglutide cardiovascular outcomes align more closely with attained dose than achieved weight loss. NPJ Cardiovasc Health. 2026;3:41. [CrossRef]
- Murugadoss K, Rajasekharan A, Malin B, et al. Building a best-in-class automated de-identification tool for electronic health records through ensemble learning. Patterns (N Y). 2021;2:100255. [CrossRef]
- Murugadoss K, Killamsetty S, Doddahonnaiah D, et al. Scaling text de-identification using locally augmented ensembles. 2024:2024.06.20.24308896. [CrossRef]
- Baggio LL, Yusta B, Mulvihill EE, et al. GLP-1 Receptor Expression Within the Human Heart. Endocrinology. 2018;159:1570–1584. [CrossRef]
Table 1.
Baseline characteristics of semaglutide and tirzepatide patients, stratified by weight-loss response. Continuous variables are presented as mean (standard deviation) and compared across all three cohorts simultaneously using one-way ANOVA. Categorical variables are presented as number (%) and compared using the Pearson chi-square test. P-values reflect omnibus testing across all three groups; a significant p-value indicates that at least one group differs from the others but does not identify which specific pair drives the difference. A two-sided p-value of <0.05 was considered statistically significant.
Table 1.
Baseline characteristics of semaglutide and tirzepatide patients, stratified by weight-loss response. Continuous variables are presented as mean (standard deviation) and compared across all three cohorts simultaneously using one-way ANOVA. Categorical variables are presented as number (%) and compared using the Pearson chi-square test. P-values reflect omnibus testing across all three groups; a significant p-value indicates that at least one group differs from the others but does not identify which specific pair drives the difference. A two-sided p-value of <0.05 was considered statistically significant.
| Characteristic | Overall cohort (n = 3,500) |
Weight-loss Super responder (n = 422) | Weight-loss moderate responder (n = 1,426) |
Weight-loss minimal responders (n = 1,652) |
P value |
|---|---|---|---|---|---|
| Age - years | 62.7 (12.3) | 61 (12) | 63.9 (11.9) | 63.9 (11.9) | <0.001 |
| Male - no. (%) | 1,910 (54.6%) | 154 (36.4%) | 790 (55.4%) | 966 (58.5%) | <0.001 |
| Race/ethnicity - no. (%) | 0.061 | ||||
| White | 3,116 (89%) | 366 (86.7%) | 1,269 (89%) | 1,481 (89.6%) | |
| African American | 229 (6.5%) | 41 (9.7%) | 92 (6.5%) | 96 (5.8%) | |
| Weight - kg | 107.6 (23.2) | 102.08 (23.1) | 107.72 (22.4) | 108.94 (23.7) | <0.001 |
| Body mass index - kg/m2 | 32.97 (4.22) | 33 (4) | 32.7 (4.3) | 33.2 (4.2) | 0.005 |
| Comorbidities - no. (%) | |||||
| Hypertension | 2789 (79.7%) | 316 (74.9%) | 1137 (79.7%) | 1336 (80.9%) | 0.024 |
| Diabetes | 2428 (69.4%) | 221 (52.4%) | 972 (68.2%) | 1235 (74.8%) | <0.001 |
| Chronic kidney disease | 1639 (46.8%) | 162 (38.4%) | 708 (49.6%) | 769 (46.5%) | <0.001 |
| Obesity | 2381 (68%) | 282 (66.8%) | 961 (67.4%) | 1138 (68.9%) | 0.575 |
| Coronary artery disease | 1544 (44.1%) | 145 (34.4%) | 642 (45.0%) | 757 (45.8%) | <0.001 |
| Heart failure | 1504 (43%) | 169 (40.0%) | 631 (44.2%) | 704 (42.6%) | 0.285 |
| Atrial fibrillation | 1033 (29.5%) | 114 (27.0%) | 454 (31.8%) | 465 (28.1%) | 0.040 |
| Time from index date to follow-up echocardiogram measurement - days | 676 (339) | 612 (319) | 649 (328) | 716 (350) | <0.001 |
Table 2.
Longitudinal echocardiographic changes in the semaglutide and tirzepatide weight-loss response groups. Echocardiographic parameters at baseline and follow-up in semaglutide or tirzepatide patients, stratified by weight-loss response group. Data are shown as mean (standard deviation) values. P-values represent within-group changes using paired t-tests. LA, left atrium; LV, left ventricle; TR Vmax, tricuspid regurgitation maximum velocity; TAPSE, tricuspid annular plane systolic excursion; RVSP, right ventricular systolic pressure.
Table 2.
Longitudinal echocardiographic changes in the semaglutide and tirzepatide weight-loss response groups. Echocardiographic parameters at baseline and follow-up in semaglutide or tirzepatide patients, stratified by weight-loss response group. Data are shown as mean (standard deviation) values. P-values represent within-group changes using paired t-tests. LA, left atrium; LV, left ventricle; TR Vmax, tricuspid regurgitation maximum velocity; TAPSE, tricuspid annular plane systolic excursion; RVSP, right ventricular systolic pressure.
| Echo parameter | Weight-loss super-responders (n = 422) |
Weight-loss moderate-responders (n = 1426) |
Weight-loss minimal-responders (n = 1652) |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Baseline | Follow -up |
P-value | Cohen’s d | Baseline | Follow -up |
P-value | Cohen’s d | Baseline | Follow -up |
P-value | Cohen’s d | |
| LA volume (mL) | 68.2 (26.8) | 66.3 (30.1) | 0.145 | -0.093 | 73.6 (31.9) | 71.0 (31.5) | 0.001 | -0.110 | 73.1 (27.7) | 72.5 (29.0) | 0.433 | -0.024 |
| Interventricular septum thickness (mm) | 11.1 (2.2) | 10.4 (2.3) | <0.001 | -0.312 | 11.1 (2.2) | 10.9 (2.2) | <0.001 | -0.148 | 11.3 (2.1) | 11.1 (2.1) | 0.002 | -0.089 |
| LV posterior wall thickness (mm) | 10.6 (1.9) | 9.9 (1.9) | <0.001 | -0.329 | 10.7 (1.9) | 10.4 (1.9) | <0.001 | -0.147 | 10.9 (1.8) | 10.7 (1.9) | <0.001 | -0.13 |
| LV relative wall thickness | 43.5 (9.1) | 42.0 (9.1) | 0.009 | -0.153 | 42.8 (9.5) | 43.2 (9.8) | 0.241 | 0.036 | 44.0 (9.6) | 43.3 (9.5) | 0.016 | -0.068 |
| LV mass (g) | 198.4 (69.6) | 176.6 (66.0) | <0.001 | -0.420 | 210.7 (74.0) | 195.8 (68.1) | <0.001 | -0.274 | 216.1 (68.4) | 204.8 (67.4) | <0.001 | -0.213 |
| LV end-diastolic diameter (mm) | 48.7 (6.8) | 47.6 (6.6) | <0.001 | -0.204 | 50.5 (7.4) | 49.2 (7.1) | <0.001 | -0.224 | 50.6 (15.3) | 49.7 (7.2) | 0.017 | -0.062 |
| LV end-systolic diameter (mm) | 32.8 (7.3) | 32.0 (7.3) | 0.027 | -0.116 | 34.3 (8.5) | 33.6 (8.0) | <0.001 | -0.11 | 33.7 (7.9) | 33.5 (7.8) | 0.255 | -0.03 |
| LV ejection fraction (%) | 58.0 (10.1) | 57.4 (10.3) | 0.187 | -0.065 | 56.8 (11.3) | 56.6 (10.7) | 0.417 | -0.022 | 57.3 (10.9) | 57.1 (10.7) | 0.201 | -0.032 |
| Mitral valve medial annulus e’ velocity | 0.1 (0.1) | 0.1 (0.0) | 0.349 | -0.052 | 0.1 (0.0) | 0.1 (0.0) | 0.050 | -0.059 | 0.1 (0.0) | 0.1 (0.0) | 0.005 | -0.077 |
| Medial E/e’ ratio | 11.8 (5.8) | 11.6 (6.2) | 0.516 | -0.037 | 12.5 (5.8) | 12.6 (6.4) | 0.934 | 0.003 | 12.7 (5.8) | 12.7 (6.6) | 0.767 | 0.008 |
| Mitral valve lateral annulus e’ velocity | 0.09 (0.03) | 0.09 (0.03) | 0.377 | -0.065 | 0.09 (0.04) | 0.09 (0.03) | 0.539 | -0.024 | 0.09 (0.02) | 0.09 (0.03) | 0.047 | -0.068 |
| Lateral E/e’ ratio | 9.1 (3.9) | 8.8 (5.6) | 0.455 | -0.056 | 9.6 (4.4) | 9.4 (4.9) | 0.247 | -0.046 | 9.8 (4.4) | 9.8 (4.6) | 0.995 | 0.000 |
| TR Vmax (m/s) | 2.5 (0.4) | 2.4 (0.4) | <0.001 | -0.268 | 2.6 (0.4) | 2.5 (0.4) | 0.021 | -0.084 | 2.6 (0.4) | 2.6 (0.4) | 0.238 | -0.041 |
| TAPSE (cm) | 20.9 (4.5) | 20.1 (5.3) | 0.051 | -0.171 | 20.2 (5.0) | 19.6 (5.2) | 0.009 | -0.118 | 20.5 (5.3) | 19.8 (5.2) | 0.001 | -0.152 |
| RVSP (mmHg) | 33.2 (12.6) | 30.4 (10.0) | 0.001 | -0.235 | 34.4 (11.3) | 33.4 (11.7) | 0.018 | -0.087 | 34.6 (11.4) | 33.7 (11.1) | 0.066 | -0.075 |
Table 3.
Comparison of longitudinal echocardiographic changes in semaglutide vs. tirzepatide weight-loss super-responders. Echocardiographic parameters at baseline and follow-up in the weight-loss super-responder group, stratified by semaglutide vs. tirzepatide therapy. Data are shown as mean (standard deviation) values. P-values represent within-group changes using paired t-tests and Cohen's d as effect size estimate. LA, left atrium; LV, left ventricle; RVSP, right ventricular systolic pressure; TAPSE, tricuspid annular plane systolic excursion; TR Vmax, tricuspid regurgitation maximum velocity.
Table 3.
Comparison of longitudinal echocardiographic changes in semaglutide vs. tirzepatide weight-loss super-responders. Echocardiographic parameters at baseline and follow-up in the weight-loss super-responder group, stratified by semaglutide vs. tirzepatide therapy. Data are shown as mean (standard deviation) values. P-values represent within-group changes using paired t-tests and Cohen's d as effect size estimate. LA, left atrium; LV, left ventricle; RVSP, right ventricular systolic pressure; TAPSE, tricuspid annular plane systolic excursion; TR Vmax, tricuspid regurgitation maximum velocity.
| Echo parameter |
Semaglutide weight-loss super-responders (n = 159) |
Tirzepatide weight-loss super-responders (n = 89) |
||||||
|---|---|---|---|---|---|---|---|---|
| Baseline | Follow -up |
P-value | Cohen’s d | Baseline | Follow -up |
P-value | Cohen’s d | |
| LA volume (mL) | 68.1 (26.3) | 66.1(28.9) | 0.352 | -0.093 | 71.6 (25.3) | 65.8 (23.5) | 0.040 | -0.288 |
| Interventricular septum thickness (mm) | 10.9 (2.2) | 10.3 (2.1) | 0.001 | -0.298 | 10.9 (1.9) | 10.8 (1.9) | 0.424 | -0.097 |
| LV posterior wall thickness (mm) | 10.4 (1.8) | 9.8 (1.8) | 0.001 | -0.299 | 10.3 (1.6) | 10.1 (1.8) | 0.244 | -0.140 |
| LV relative wall thickness (mm) |
43.5 (9.3) | 41.8 (9.1) | 0.070 | -0.167 | 42 (7.2) | 42.4 (9) | 0.696 | 0.047 |
| LV mass (g) | 196 (67.9) | 177.6 (64.8) | <0.001 | -0.389 | 197.3 (53.4) | 184.2 (59.7) | 0.032 | -0.263 |
| LV end-diastolic diameter (mm) | 49 (6.4) | 47.9 (6.3) | 0.008 | -0.228 | 50.1 (5.5) | 48.7 (6.1) | 0.015 | -0.284 |
| LV end-systolic diameter (mm) | 32.9 (6.7) | 32.7 (7.1) | 0.325 | -0.086 | 33.8 (6.6) | 33.1 (7.2) | 0.182 | -0.160 |
| LV ejection fraction | 59.1 (8.9) | 58.3 (10.4) | 0.305 | -0.083 | 58.5 (10) | 56.9 (10.7) | 0.028 | -0.237 |
| Mitral valve medial annulus e’ velocity | 0.075 (0.05) | 0.067 (0.03) | 0.063 | -0.161 | 0.074 (0.02) | 0.078 (0.08) | 0.719 | 0.040 |
| Medial E/e’ ratio | 11.9 (5.7) | 12.3 (6.2) | 0.427 | 0.070 | 11.7 (6) | 11.8 (7.7) | 0.843 | 0.022 |
| Mitral valve lateral annulus e’ velocity | 0.092 (0.03) | 0.089 (0.03) | 0.192 | -0.142 | 0.096 (0.03) | 0.092 (0.03) | 0.189 | -0.188 |
| Lateral E/e’ ratio | 9 (4) | 9 (5.1) | 0.998 | 0.000 | 9.1 (4.3) | 9.1 (6.1) | 0.977 | -0.004 |
| TR Vmax (m/s) | 2.6 (0.5) | 2.5 (0.5) | 0.131 | -0.161 | 2.5 (0.5) | 2.4 (0.4) | 0.044 | -0.286 |
| TAPSE (cm) | 21.2 (4.8) | 20.2 (5.4) | 0.131 | -0.196 | 20.8 (4.4) | 19 (4.4) | 0.038 | -0.411 |
| RVSP (mmHg) | 33.9 (12) | 31.7 (11.1) | 0.078 | -0.189 | 32.4 (13.3) | 29.6 (10.3) | 0.094 | -0.239 |
Table 4.
Comparison of longitudinal echocardiographic changes in semaglutide and tirzepatide patients vs. propensity-matched patients on non-incretin weight-loss medications. Echocardiographic parameters at baseline and follow-up in patients on semaglutide or tirzepatide with >5% weight loss propensity-matched with patients on non-incretin weight-loss medications. Data are shown as mean (standard deviation) values. P-values represent within-group changes using paired t-tests and Cohen's d as effect size estimate. LA, left atrium; LV, left ventricle; TR Vmax, tricuspid regurgitation maximum velocity; TAPSE, tricuspid annular plane systolic excursion; RVSP, right ventricular systolic pressure.
Table 4.
Comparison of longitudinal echocardiographic changes in semaglutide and tirzepatide patients vs. propensity-matched patients on non-incretin weight-loss medications. Echocardiographic parameters at baseline and follow-up in patients on semaglutide or tirzepatide with >5% weight loss propensity-matched with patients on non-incretin weight-loss medications. Data are shown as mean (standard deviation) values. P-values represent within-group changes using paired t-tests and Cohen's d as effect size estimate. LA, left atrium; LV, left ventricle; TR Vmax, tricuspid regurgitation maximum velocity; TAPSE, tricuspid annular plane systolic excursion; RVSP, right ventricular systolic pressure.
| Echo parameter |
GLP-1/GIP receptor agonist >5% weight loss responder (n = 118) |
Non-incretin weight-loss medication >5% weight-loss responder (n = 118) |
||||||
|---|---|---|---|---|---|---|---|---|
| Baseline | Follow -up |
P-value | Cohen’s d | Baseline | Follow -up |
P-value | Cohen’s d | |
| LA volume (mL) | 71.7 (35.0) | 68.5 (32.9) | 0.246 | -0.137 | 65.1 (28.8) | 69.3 (32.4) | 0.108 | 0.175 |
| Interventricular septum thickness (mm) | 10.6 (2.0) | 10.2 (1.8) | 0.030 | -0.229 | 10.6 (2.1) | 10.4 (1.9) | 0.304 | -0.101 |
| LV posterior wall thickness (mm) | 10.4 (1.7) | 9.6 (1.4) | <0.001 | -0.443 | 9.8 (1.7) | 9.8 (2.0) | 0.954 | -0.008 |
| LV relative wall thickness (mm) | 42.8 (9.3) | 40.9 (7.5) | 0.046 | -0.209 | 40.9 (9.0) | 42.9 (10.5) | 0.255 | 0.152 |
| LV mass (g) | 194.2 (58.3) | 171.7 (56.8) | <0.001 | -0.499 | 186.1 (53.0) | 179.1 (61.4) | 0.157 | -0.139 |
| LV end-diastolic diameter (mm) | 49.6 (7.0) | 48.2 (6.8) | 0.006 | -0.264 | 49.2 (4.9) | 47.0 (5.7) | 0.002 | -0.402 |
| LV end-systolic diameter (mm) | 33.7 (7.0) | 32.6 (6.9) | 0.021 | -0.229 | 32.3 (5.3) | 31.0 (5.8) | 0.050 | -0.254 |
| LV ejection fraction (%) | 57.9 (8.9) | 58.3 (9.4) | 0.522 | 0.059 | 60.8 (8.8) | 59.9 (8.5) | 0.437 | -0.095 |
| Medial e’ velocity (cm/s) | 0.085 (0.08) | 0.076 (0.02) | 0.289 | -0.105 | 0.079 (0.04) | 0.072 (0.02) | 0.318 | -0.101 |
| Medial E/e’ ratio | 11.2 (4.8) | 11.1 (6.2) | 0.808 | -0.024 | 11.1 (4.5) | 11.7 (3.9) | 0.245 | 0.153 |
| Lateral e’ velocity (cm/s) | 0.100 (0.03) | 0.096 (0.03) | 0.440 | -0.116 | 0.089 (0.024) | 0.098 (0.026) | 0.027 | 0.345 |
| Lateral E/e’ ratio | 8.4 (3.3) | 8.6 (5.1) | 0.766 | 0.045 | 8.8 (3.5) | 8.3 (2.7) | 0.297 | -0.205 |
| TR Vmax (m/s) | 2.52 (0.44) | 2.51 (0.36) | 0.804 | -0.032 | 2.54 (0.37) | 2.51 (0.31) | 0.629 | -0.085 |
| TAPSE (cm) | 19.8 (4.8) | 19.8 (4.8) | 1.000 | 0.000 | 21.7 (4.1) | 21.4 (4.1) | 0.739 | -0.054 |
| RVSP (mmHg) | 31.9 (9.4) | 31.9 (9.3) | 1.000 | 0.000 | 33.3 (10.6) | 33.9 (10.3) | 0.668 | 0.055 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2026 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.
