ABSTRACT
Background
Echocardiogram is commonly used to evaluate cardiac remodeling in hypertension (HTN). However, study on echocardiographic phenotypes and their prognostic implications in HTN is limited.
Objective
We aimed to evaluate the prognostic implications echocardiographic phenotypes in community hypertensive patients.
Method
A total of 1881 community hypertensive patients without overt cardiovascular disease and severe renal disease (mean age 62.8 years, women 57.9%) were included. Using Two-Step cluster analysis with four conventional echocardiographic variables, two clusters with distinct echocardiographic phenotypes were identified.
Result
The Cluster 1 (namely “mild-remodeling” HTN; n = 1492) had low prevalence of enlarged left atrium (LA; 0.9%) and left ventricular hypertrophy (LVH; 16.2%) and better LV diastolic function. They were younger and more likely to be men and had lower comorbid burden. The Cluster 2 (namely “severe-remodeling” HTN; n = 389) had higher prevalence of enlarged LA (26.0%) and LVH (83.0%) and worse LV diastolic function. They were older and more likely to be women and had higher comorbid burden. After a median follow-up of 4.2 years, compared to the Cluster 1, the Cluster 2 had higher incidence of cardiovascular (4.1% vs 1.7%; P = .006) and all-cause (9.8% vs 4.8%; P < .001) death, with adjusted hazard ratio of 2.80 (95% CI 1.39–5.62; P = .004) and 2.04 (95% CI 1.32–3.14; P < .001) respectively.
Conclusion
These findings indicate that the conventional echocardiographic variables-based algorithm could help identify asymptomatic community hypertensive patients at risk for cardiovascular and all-cause death. Further studies are needed to develop and validate phenotype-specific prevention and intervention strategies in HTN.
Introduction
Hypertension (HTN) is a major public health issue worldwide including China (Citation1,Citation2). As estimated, there are nearly 244.5 million (about 1 in 4) adults have HTN in China (Citation3), which accounts for nearly one-fifth of overall hypertensive cases globally. Notably, blood pressure (BP) elevation causes cardiac remodeling (Citation4,Citation5), which is independently associated with cardiovascular and all-cause death (Citation6–8). Therefore, screening for cardiac remodeling could help identify asymptomatic hypertensive patients at risk for cardiovascular events, which may help guide prevention and intervention strategies.
Diabetes mellitus (DM) commonly coexists with HTN, and many cardiac remodeling found in hypertensive patients are analogous to those observed in diabetic patients (Citation9–11). Furthermore, both our and other studies have shown that age, sex, obesity, and DM could influence cardiac remodeling in hypertensive patients (Citation11–16), and the specific contribution of these factors is not fully elucidated yet, as is their joint effect on cardiac remodeling in HTN.
Cluster analysis, an unsupervised machine learning algorithm, has been broadly used in research to identify specific phenotype of disease with heterogeneous characteristic (Citation17). Prior studies have shown that cluster analysis of echocardiographic variables for assessing cardiac structure and function could identify unique echocardiographic phenotypes, which display differential clinical profiles and outcomes (Citation18–20). Despite echocardiogram has been commonly used in daily clinical practice to evaluate cardiac remodeling in HTN, to our knowledge, study on machine learning-derived echocardiographic phenotypes and their prognostic implication in HTN is limited. We hypothesized that in community hypertensive patients without overt cardiovascular disease and severe renal disease, leveraging cluster analysis, we were able to identify unique echocardiographic phenotypes based on conventional echocardiographic variables, and there were significant differences in the clinical profiles and outcomes between these phenotypes.
Methods
Study participants
Hypertensive patients from Liaobu County (Dongguan, China), who have attended the local government-sponsored annual health examination including echocardiographic examination in 2014, were screened for the eligibility of this study. The detailed information of this study has been described previously (Citation14). Exclusion criteria were prior history of coronary heart disease and stroke, severe renal disease as defined by estimated glomerular filtration rate (eGFR) < 30 ml/min/1.73 m2, missed eGFR value, and incomplete echocardiographic variables. This study was approved by the Clinical Research Ethics Committee of Guangdong Provincial People’s Hospital and the Liaobu County Health Center. Informed consent was obtained before enrollment.
Data collection and study variables
Standardized questionnaire was used to collect data on demographic, prior medical history, and current medical therapy by trained staffs during health examination. Body weight was measured without wearing heavy cloths and height was measured without wearing shoes. Body mass index (BMI) was calculated as body weight in kilograms divided by height in squared meters, with a BMI ≥28 kg/m2 was defined as obesity (Citation21). BP was measured according to the China Hypertension Guideline (Citation22), and at least two BP measurements with 1-minute interval was performed using Omron HEM-7051 device (Omron HealthCare, Kyoto, Japan). The average value of two BP readings was recorded. Fasting venous blood was used to measure lipid panel, fasting plasma glucose (FPG), and serum creatinine level, which was used to calculate eGFR using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation (Citation23), with an eGFR <60 ml/min/1.73 m2 was defined as chronic kidney disease (CKD).
Echocardiographic variables
Transthoracic echocardiographic examination was performed at rest using a Vivid S6 M4S-RS Probe (GE Ving-Med) interfaced with a 2.5- to 3.5-MHz phased-array probe. All the measurements were performed by experienced sonographers according to the American Society Echocardiography/European Association of Cardiovascular Imaging (ASE/EACVI) Guideline (Citation24). Conventional echocardiographic variables for assessing cardiac structure and function were used. Left ventricular end diastolic and systolic diameter (LVEDD and LVESD) and left atrial diameter (LAD) was indexed to height and left ventricular mass (LVM), which was estimated using linear echocardiographic dimensions (Citation25), was indexed to height2.7. Enlarged left atrium (LA) was defined as LA anteroposterior diameter >4.0 cm for men and >3.8 cm for women, or LAD/height >2.61 cm/m for both sexes; left ventricular hypertrophy (LVH) was defined as LVM/height2.7 >48 g/m2.7 for men and >49 g/m2.7 for women; and concentric remodeling (CR) was defined as relative wall thickness (RWT) > 0.42 for both sexes. Based on presence of CR and LVH, left ventricular geometry was divided into normal, CR, concentric LVH and eccentric LVH, respectively.
Cluster analysis
We used Two-Step cluster analysis to explore the echocardiographic phenotypes, and the four conventional echocardiographic variables were as follows: RWT, LVM/height2.7, LAD/height and E/A ratio, which reflect LV remodeling and diastolic function. In brief, the Two-Step cluster analysis used Schwarz’s Bayesian information criterion (BIC) to determine the number of clusters and Log-likelihood as the distance measure (IBM SPSS Modeler 18.3 Algorithms Guide reference. Chicago, IL, USA: IBM; 2021.). In the first step (pre-clustering), a BIRCH (Balanced Iterative Reducing and Clustering Hierarchies) algorithm was used to pre-cluster the cases, and in the second step (clustering), these pre-clusters were joined to clusters using an agglomerative hierarchical algorithm. Such a technique has several advantages such as deciding the number of clusters based on a statistical measure of fit (BIC or, optionally, AIC [Akaike Information Criterion]) rather than on an arbitrary choice and analyzing atypical values such as outliers (Citation26). In addition, the Two-Step cluster analysis has been recognized as one of the most reliable techniques in terms of the number of novel groups detected, classification probability of individuals into novel groups, and reproducibility of findings on clinical and other types of data (Citation26). All these analyses were conducted by author Zhiqiang Nie using a similar method he has reported previously (Citation27).
Study outcome
The primary outcome was cardiovascular death, and the causes included myocardial infarction, stroke, and heart failure. The secondary outcome was all-cause death, including cardiovascular and non-cardiovascular death, and the latter was broadly divided into cancer- and non-cancer-related death. All these events were obtained from the database which is linked to the Dongguan Medical Insurance Bureau and Liaobu Community Health Center, both are the government agencies to register the vital status and the causes of death of the residents. Study participants were censored at the time of event occurred or till the end of December 31, 2018.
Analytic plan
We first separated study participants into HTN alone and HTN plus DM groups. Clinical profiles, echocardiographic variables and outcomes were compared between these two clinical groups. In the second step, study participants were divided into different cluster groups based on cluster analysis. Clinical profiles, echocardiographic variables and outcomes were compared between these cluster groups.
Statistical analysis
Continuous variables were presented as mean (standard deviation; SD) and categorical variables were presented as frequency (proportion). Clinical profiles, echocardiographic variables and outcomes were compared using Student t test or Chi-squared test as appropriate. Adjusted cumulative incidence curves were plotted using the Kaplan–Meier method for cardiovascular and all-cause death, using the log-rank test to assess the significance for groups comparison. Cox proportional hazards regression models were constructed to compute hazard ratios (HR) and 95% confidence intervals (CI), with adjustment for age, sex, BMI, heart rate, smoking, eGFR, dyslipidemia, antihypertensive drug, and lipid-lowering medication. The proportional hazards assumption was examined using Schoenfeld residuals, and there was no evidence that the proportional hazards assumption was violated. Two-sided P-value <.05 was defined as statistical analysis. All the analysis was performed using the IBM SPSS software (v 26.0) and the R platform (v4.1).
Results
Among the 2445 community hypertensive patients who had echocardiographic examination, after excluding those with coronary heart disease, stroke, eGFR <30 ml/min/1.73 m2, missed eGFR value and incomplete echocardiographic variables, a total of 1881 hypertensive patients were included (). The mean age was 62.8 years, women were 57.9%, and the prevalence of DM was 21.0%.
Figure 1. Study flowchart.
Cluster groups
The cluster analysis identified two clusters with differential echocardiographic phenotypes. As is shown in , the two clusters were phenotypically distinct with regard to the four echocardiographic variables. The Cluster 1, including 1492 (79.3%) of the 1881 clustered patients, had a lower RWT, LVM/height2.7 and LAD/height while a higher E/A ratio. They had a lower prevalence of enlarged LA and LVH and better LV diastolic function and was thus labeled as “mild-remodeling” HTN. In contrast, the Cluster 2, including 389 patients (20.7%), had a higher RWT, LVM/height2.7 and LAD/height while a lower E/A ratio. They had a high prevalence of enlarged LA and LVH and poorer LV diastolic function and was thus labeled as “severe-remodeling” HTN. Radar charts illustrated the significant differences in echocardiographic variables between these two clusters (). In addition, radar chart showed that there were differences in the clinical profiles between these two clusters ().
Figure 2. Cluster analysis.
General characteristics in the clinical and cluster groups
General characteristics of patients in the HTN alone and HTN plus DM groups are summarized in . Patients in the HTN plus DM group had higher BMI and triglyceride, and higher prevalence of obesity and dyslipidemia. In addition, they were more likely to use lipid-lowering medication, angiotensin-converting enzyme inhibitor and calcium channel blocker (CCB). Compared to patients in the Cluster 1, those in the Cluster 2 were older, more likely to be women, and had higher BMI, SBP and heart rate. In addition, they had higher prevalence of obesity and CKD, and they were more likely to use lipid-lowering medication, CCB and diuretic.
Table 1. General characteristics.
Echocardiographic variables in the clinical and cluster groups
Differences in interventricular septum [IVS], LV posterior wall [LVPW], RWT, LAD and LAD/height were observed between the two clinical groups (). Nonetheless, as is shown in , there was significant overlap of individual value of these echocardiographic variables. In contrast, there were significant differences in the overall echocardiographic variables between these two cluster groups, and there were less overlapped of individual value of these echocardiographic variables.
Figure 3. Echocardiographic variables.
Table 2. Echocardiographic variables.
Outcomes in the clinical and cluster groups
After a median follow-up of 4.2 (interquartile range: 4.0–4.3) years, the incidence of cardiovascular death in the HTN plus DM and HTN alone groups were 2.5% and 2.1% respectively, with adjusted HR was 1.29 (95% CI 0.62–2.69; P = .50; ). While patients in the HTN plus DM group had a higher incidence of all-cause death (8.4% vs 5.2%), with adjusted HR was 1.56 (95% CI 1.02–2.39; P = .04). Compared to the Cluster 1, patients in the Cluster 2 group had a higher incidence of cardiovascular (4.1% vs 1.7%) and all-cause (9.8% vs 4.8%) death, with adjusted HR of 2.80 (95% CI 1.39–5.62; P = .004) and 2.04 (95% CI 1.32–3.14; P < .001), respectively. Patients in the HTN plus DM group had a higher adjusted cumulative incidence of all-cause death and there was no difference in the adjusted cumulative incidence of cardiovascular death (); while patients in the Cluster 2 had a higher adjusted cumulative incidence of cardiovascular and all-cause death than those in the Cluster 1 ().
Figure 4. Kaplan-Meier curve.
Table 3. Cardiovascular and all-cause death.
Discussion
Leveraging cluster analysis, we are able to identify two unique echocardiographic phenotypes with differential clinical profiles, cardiac remodeling and outcomes. In specific, the Cluster 1, namely “mild-remodeling” HTN, were younger and more likely to be men, and had lower comorbid burden and lower prevalence of enlarged LA and LVH and better LV diastolic function. In contrast, the Cluster 2, namely “severe-remodeling” HTN, were older and more likely to be women, and had higher comorbid burden and higher prevalence of enlarged LA and LVH and worse LV diastolic function. Importantly, the “severe-remodeling” HTN phenotype was associated with a higher risk of cardiovascular and all-cause death. These findings indicate that the conventional echocardiographic variables-based algorithm could help identify asymptomatic community hypertensive patients at risk for cardiovascular and all-cause death. Further studies are needed to develop and validate the phenotype-specific prevention and intervention strategies in HTN.
It has been well demonstrated that presence of cardiac remodeling (such as enlarged LA and LVH) portends a high mortality risk in hypertensive patients (Citation6–8). Notably, HTN is an independent risk factor for cardiac remodeling. Nevertheless, both our and other studies have indicated that cardiac remodeling associated with HTN could be confounded by other factors such as age, sex, obesity and DM (Citation11–16). Therefore, to quantify and interpret the contribution of HTN alone to cardiac remodeling is challenging. Indeed, as is shown in this study, although there were statistical differences in the average values of echocardiographic variables between the HTN alone and HTN plus DM groups, individual value of these variables was significantly overlapped. Furthermore, there were no differences in the prevalence of enlarged LA and LVH, which might explain no differences in cardiovascular death between these two clinical groups. These findings suggest that traditional statistical analysis, which is based on a priori hypothesis, might not be able to discriminate asymptomatic hypertensive patients at risk for cardiovascular death.
Cluster analysis has been increasingly used in research recently, and the goal is to try to learn the intrinsic structure within data to derive novel phenotype groups of a disease or clinical syndrome (Citation17). These findings subsequently can be potentially used to guide phenotype-specific prevention and intervention strategies (Citation17). For example, Katz et al used cluster analysis to identify two distinct hypertensive groups that were associated with different myocardial substrate for heart failure with preserved ejection fraction (HFpEF), based on which targeted therapies could be developed to prevent HFpEF (Citation28). Two-Step cluster analysis is one of the common used algorithms of cluster analysis and the advantages have been described above (Citation26).
To our knowledge, this should be the first few studies to apply Two-Step cluster analysis to explore echocardiographic phenotypes and their prognostic implications in asymptomatic community hypertensive patients. Based on four conventional echocardiographic variables, two clusters with distinct echocardiographic phenotypes were identified. In specific, the Cluster 1 was characterized by less LV wall thickness, smaller LV and LA diameter, lower RWT, higher E/A ratio, and lower prevalence of enlarged LA and LVH. They had better clinical profiles as reflected by younger age and lower prevalence of obesity and CKD. These together might contribute to their favorable outcomes. In contrast, the Cluster 2 was characterized by greater LV wall thickness, larger LV and LA diameter, higher RWT, lower E/A ratio, and higher prevalence of enlarged LA and LVH. In addition, they were older and more likely to be women, and had a higher prevalence of obesity and CKD. These unfavorable characteristics might contribute to their worse outcomes. Taken together, these results support our hypothesis that leveraging cluster analysis, we are able to identify unique echocardiographic phenotypes in asymptomatic hypertensive patients. In addition, this study demonstrates the prognostic implications of the conventional echocardiographic variables-based algorithm, and it is important to improve the assessment of cardiac structure and function in HTN. Furthermore, these findings were in line with prior studies of the association between cardiac remodeling with mortality risk in HTN (Citation6–8).
Of note, both the European and American hypertension guidelines recommend to screen and monitor target organ damage such as left ventricular hypertrophy or diastolic dysfunction (Citation29,Citation30). Presence of cardiac structural and function changes may justify intensive treatment of HTN and close follow-up. Findings of current study support these recommendations and suggest that machine-learning algorithm (cluster analysis) could help better identify specific echocardiography phenotypes, which can guide clinical managements.
Both epidemiologic studies and clinical trials have consistently demonstrated that HFpEF populations are more likely to be older women with high comorbid burden such as obesity and CKD (Citation31–35). In addition, they are more likely to have structural heart disease such as enlarged LA, LVH and reduced LV diastolic function. Notably, clinical profiles and echocardiographic features of hypertensive patients in the Cluster 2 were in line with prior reports, suggesting that the Cluster 2 might gather a population group that is at risk for developing HFpEF. In addition, these results also highlight the potential value of combining conventional echocardiographic variables and the state-of-the-art statistical technique in improving HTN management.
Despite some important findings of this study, there are some limitations deserve attention. First, hypertensive patients with overt cardiovascular disease and severe renal disease were excluded, therefore, these findings could not be extrapolated to patients with these diseases. Second, a small proportion of patients with incomplete echocardiographic variables were excluded, which might result in biases of these findings. Third, this study only used four conventional echocardiographic variables to explore echocardiographic phenotypes, and whether including other variables such as LV global longitudinal strain and E/e’ ratio would help identify novel cluster groups deserves further elucidation. Fourth, in the current study, the prevalence of atrial fibrillation was unknown which might influence the association between echocardiographic phenotypes and outcomes. Fifth, we did not validate these findings in the other populations and it should be cautious when extrapolating these findings to other population groups.
Conclusion
Leveraging cluster analysis, we herein identify two cluster groups with distinct echocardiographic phenotypes in asymptomatic community hypertensive patients, and these two clusters display differential clinical profiles and outcomes. If confirmed in other studies, we believe that the conventional echocardiographic variables-based algorithm could help guide development of the phenotype-specific prevention and intervention strategies in HTN in the future.
Acknowledgments
We thank the participants and all health staffs in Liaobu County for assistance with data collection.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Additional information
Funding
References
- Mills KT, Stefanescu A, He J. The global epidemiology of hypertension. Nat Rev Nephrol. 2020;16(4):223–9. doi:10.1038/s41581-019-0244-2.
- Zhou B, Carrillo-Larco RM, Danaei G, Riley LM, Paciorek CJ, Stevens GA, Gregg EW, Bennett JE, Solomon B, Singleton RK. Worldwide trends in hypertension prevalence and progress in treatment and control from 1990 to 2019: a pooled analysis of 1201 population-representative studies with 104 million participants. Lancet. 2021;398(10304):957–80. doi:10.1016/s0140-6736(21)01330-1.
- Wang Z, Chen Z, Zhang L, Wang X, Hao G, Zhang Z, Shao L, Tian Y, Dong Y, Zheng C, et al. Status of hypertension in China: results from the China hypertension survey, 2012-2015. Circulation. 2018;137(22):2344–56. doi:10.1161/circulationaha.117.032380.
- Vasan RS, Song RJ, Xanthakis V, Beiser A, DeCarli C, Mitchell GF, Seshadri S. Hypertension-mediated organ damage: prevalence, correlates, and prognosis in the community. Hypertension. 2022;79(3):505–15. doi:10.1161/hypertensionaha.121.18502.
- Tran AH, Flynn JT, Becker RC, Daniels SR, Falkner BE, Ferguson M, Hanevold CD, Hooper SR, Ingelfinger JR, Lande MB, et al. Subclinical systolic and diastolic dysfunction is evident in youth with elevated blood pressure. Hypertension. 2020;75(6):1551–56. doi:10.1161/hypertensionaha.119.14682.
- Bang CN, Soliman EZ, Simpson LM, Davis BR, Devereux RB, Okin PM. Electrocardiographic left ventricular hypertrophy predicts cardiovascular morbidity and mortality in hypertensive patients: the allhat study. Am J Hypertens. 2017;30(9):914–22. doi:10.1093/ajh/hpx067.
- Okin PM, Hille DA, Kjeldsen SE, Devereux RB. Combining ECG criteria for left ventricular hypertrophy improves risk prediction in patients with hypertension. J Am Heart Assoc. 2017;6(11). doi:10.1161/jaha.117.007564.
- Chirinos JA, Sardana M, Ansari B, Satija V, Kuriakose D, Edelstein I, Oldland G, Miller R, Gaddam S, Lee J, et al. Left atrial phasic function by cardiac magnetic resonance feature tracking is a strong predictor of incident cardiovascular events. Circ Cardiovasc Imaging. 2018;11(12):e007512. doi:10.1161/circimaging.117.007512.
- Galderisi M. Diastolic dysfunction and diabetic cardiomyopathy: evaluation by Doppler echocardiography. J Am Coll Cardiol. 2006;48(8):1548–51. doi:10.1016/j.jacc.2006.07.033.
- Fang ZY, Yuda S, Anderson V, Short L, Case C, Marwick TH. Echocardiographic detection of early diabetic myocardial disease. J Am Coll Cardiol. 2003;41(4):611–17. doi:10.1016/s0735-1097(02)02869-3.
- von Jeinsen B, Vasan RS, McManus DD, Mitchell GF, Cheng S, Xanthakis V. Joint influences of obesity, diabetes, and hypertension on indices of ventricular remodeling: findings from the community-based Framingham heart study. PLoS One. 2020;15(12):e0243199. doi:10.1371/journal.pone.0243199.
- Cai A, Zhou D, Liu L, Zhou Y, Tang S, Feng Y. Age-related alterations in cardiac and arterial structure and function in hypertensive women and men. J Clin Hypertens. 2021;23(7):1322–34. doi:10.1111/jch.14262.
- Cai A, Liu L, Zhou D, Zhou Y, Tang S, Feng Y. The patterns of left ventricular alteration by adipose tissue distribution: implication for heart failure prevention. ESC Heart Fail. 2021;8(4):3093–105. doi:10.1002/ehf2.13415.
- Cai A, Liu L, Zhou D, Tang S, Zhou Y, Feng Y. Influences of achieved SBP on age and sex-related left ventricular structural alteration in community hypertensive populations. J Hypertens. 2022;40(6):1170–78. doi:10.1097/hjh.0000000000003127.
- Tadic M, Cuspidi C, Celic V, Ivanovic B, Pencic B, Grassi G. The influence of sex on left ventricular strain in hypertensive population. J Hypertens. 2019;37(1):50–56. doi:10.1097/hjh.0000000000001838.
- Hoshida S, Shinoda Y, Ikeoka K, Fukuoka H, Inui H, Watanabe T. Age- and sex-related differences in diastolic function and cardiac dimensions in a hypertensive population. ESC Heart Fail. 2016;3(4):270–77. doi:10.1002/ehf2.12097.
- Quer G, Arnaout R, Henne M, Arnaout R. Machine learning and the future of cardiovascular care: JACC state-of-the-art review. J Am Coll Cardiol. 2021;77(3):300–13. doi:10.1016/j.jacc.2020.11.030.
- Lancaster MC, Salem Omar AM, Narula S, Kulkarni H, Narula J, Sengupta PP. Phenotypic clustering of left ventricular diastolic function parameters: patterns and prognostic relevance. JACC Cardiovasc Imaging. 2019;12(7):1149–61. doi:10.1016/j.jcmg.2018.02.005.
- Kobayashi M, Huttin O, Magnusson M, Ferreira JP, Bozec E, Huby A-C, Preud’homme G, Duarte K, Lamiral Z, Dalleau K, et al. Machine learning-derived echocardiographic phenotypes predict heart failure incidence in asymptomatic individuals. JACC Cardiovasc Imaging. 2022;15(2):193–208. doi:10.1016/j.jcmg.2021.07.004.
- Ernande L, Audureau E, Jellis CL, Bergerot C, Henegar C, Sawaki D, Czibik G, Volpi C, Canoui-Poitrine F, Thibault H, et al. Clinical implications of echocardiographic phenotypes of patients with diabetes mellitus. J Am Coll Cardiol. 2017;70(14):1704–16. doi:10.1016/j.jacc.2017.07.792.
- Zhou BF. Predictive values of body mass index and waist circumference for risk factors of certain related diseases in Chinese adults--study on optimal cut-off points of body mass index and waist circumference in Chinese adults. Biomed Environ Sci. 2002;15(1):83–96.
- 2018 Chinese guidelines for prevention and treatment of hypertension-a report of the revision committee of Chinese guidelines for prevention and treatment of hypertension. J Geriatr Cardiol. 2019;16:182–241. doi:10.11909/j.issn.1671-5411.2019.03.014.
- Levey AS, Stevens LA, Schmid CH, Zhang Y(, Castro AF, Feldman HI, Kusek JW, Eggers P, Van Lente F, Greene T, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009;150(9):604–12. doi:10.7326/0003-4819-150-9-200905050-00006.
- Lang RM, Badano LP, Mor-Avi V, Afilalo J, Armstrong A, Ernande L, Flachskampf FA, Foster E, Goldstein SA, Kuznetsova T, et al. Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American society of echocardiography and the European association of cardiovascular imaging. Eur Heart J Cardiovasc Imaging. 2015;16(3):233–70. doi:10.1093/ehjci/jev014.
- Marwick TH, Gillebert TC, Aurigemma G, Chirinos J, Derumeaux G, Galderisi M, Gottdiener J, Haluska B, Ofili E, Segers P, et al. Recommendations on the use of echocardiography in adult hypertension: a report from the European Association of Cardiovascular Imaging (EACVI) and the American Society of Echocardiography (ASE)†. Eur Heart J Cardiovasc Imaging. 2015;16(6):577–605. doi:10.1093/ehjci/jev076.
- Kent P, Jensen RK, Kongsted A. A comparison of three clustering methods for finding subgroups in MRI, SMS or clinical data: SPSS TwoStep cluster analysis, latent gold and SNOB. BMC Med Res Methodol. 2014;14(1):113. doi:10.1186/1471-2288-14-113.
- Yang DY, Nie ZQ, Liao LZ, Zhang S-Z, Zhou H-M, Sun X-T, Zhong X-B, Du Z-M, Zhuang X-D, Liao X-X, et al. Phenomapping of subgroups in hypertensive patients using unsupervised data-driven cluster analysis: An exploratory study of the SPRINT trial. Eur J Prev Cardiol. 2019;26(16):1693–706. doi:10.1177/2047487319856733.
- Katz DH, Deo RC, Aguilar FG, Selvaraj S, Martinez EE, Beussink-Nelson L, Kim K-YA, Peng J, Irvin MR, Tiwari H, et al. Phenomapping for the identification of hypertensive patients with the myocardial substrate for heart failure with preserved ejection fraction. J Cardiovasc Transl Res. 2017;10(3):275–84. doi:10.1007/s12265-017-9739-z.
- Whelton PK, Carey RM, Aronow WS, Casey DE, Collins KJ, Dennison Himmelfarb C, DePalma SM, Gidding S, Jamerson KA, Jones DW, et al. 2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA guideline for the prevention, detection, evaluation, and management of high blood pressure in adults: A report of the American college of cardiology/American heart association task force on clinical practice guidelines. Hypertension. 2018;71(6):e13–e115. doi:10.1161/hyp.0000000000000065.
- Williams B, Mancia G, Spiering W, Agabiti Rosei E, Azizi M, Burnier M, Clement DL, Coca A, de Simone G, Dominiczak A, et al. 2018 ESC/ESH guidelines for the management of arterial hypertension. Eur Heart J. 2018;39(33):3021–104. doi:10.1093/eurheartj/ehy339.
- Tromp J, Shen L, Jhund PS, Anand IS, Carson PE, Desai AS, Granger CB, Komajda M, McKelvie RS, Pfeffer MA, et al. Age-related characteristics and outcomes of patients with heart failure with preserved ejection fraction. J Am Coll Cardiol. 2019;74(5):601–12. doi:10.1016/j.jacc.2019.05.052.
- Cohen JB, Schrauben SJ, Zhao L, Basso MD, Cvijic ME, Li Z, Yarde M, Wang Z, Bhattacharya PT, Chirinos DA, et al. Clinical phenogroups in heart failure with preserved ejection fraction: detailed phenotypes, prognosis, and response to spironolactone. JACC Heart Fail. 2020;8(3):172–84. doi:10.1016/j.jchf.2019.09.009.
- Sotomi Y, Hikoso S, Nakatani D, Mizuno H, Okada K, Dohi T, Kitamura T, Sunaga A, Kida H, Oeun B, et al. Sex differences in heart failure with preserved ejection fraction. J Am Heart Assoc. 2021;10(5):e018574. doi:10.1161/jaha.120.018574.
- Pitt B, Pfeffer MA, Assmann SF, Boineau R, Anand IS, Claggett B, Clausell N, Desai AS, Diaz R, Fleg JL, et al. Spironolactone for heart failure with preserved ejection fraction. N Engl J Med. 2014;370(15):1383–92. doi:10.1056/NEJMoa1313731.
- Solomon SD, McMurray JJV, Anand IS, Ge J, Lam CSP, Maggioni AP, Martinez F, Packer M, Pfeffer MA, Pieske B, et al. Angiotensin–neprilysin inhibition in heart failure with preserved ejection fraction. N Engl J Med. 2019;381(17):1609–20. doi:10.1056/NEJMoa1908655.