Prediction of incident heart failure in type 2 diabetes mellitus: Rationale and design of the DM‐HEART study

ABSTRACT

Preclinical heart failure (HF) (stages A and B) is common but easily overlooked in asymptomatic patients with type 2 diabetes (T2D). There is a paucity of evidence regarding how coexisting risk factors contribute to the development of HF in asymptomatic Asian patients with T2D, and there is a lack of guidance on the early identification of asymptomatic patients to help cost‐effectively allocate preventive therapies. The Diabetes Mellitus Heart Failure Events Analysis and Risk Tracking (DM‐HEART) trial, a prospective, multicenter, longitudinal cohort study, recruited asymptomatic patients with T2D and no prior HF between January 2021 and June 2024, with a minimum follow‐up period of 12 months. The primary outcome was incident HF, defined as new‐onset HF (requiring either hospitalization or an urgent care visit). Secondary outcomes are measured as 3‐point major adverse cardiovascular events, comprised of cardiovascular death, non‐fatal myocardial infarction, or non‐fatal stroke. DM‐HEART will provide epidemiological data on the incidence of HF and assess the prognostic utility of clinical parameters, including N‐terminal pro‐B‐type natriuretic peptide (NT‐proBNP), for predicting incident HF events among asymptomatic patients with T2D in preclinical stages of HF. Furthermore, we aim to determine the optimal NT‐proBNP cut‐off value for predicting incident HF events in East Asian patients with diabetes.

INTRODUCTION

Diabetes is a global pandemic affecting over 10% of adults worldwide. In 2024, the number of adults with diabetes is projected to exceed half a billion, potentially reaching nearly one billion by 2050 ¹ . Meanwhile, heart failure (HF) has been recognized as an underappreciated complication of diabetes, with a prevalence of up to 22% among individuals with diabetes, and which is expected to rise due to advanced treatments and the overall longer life expectancy of the general population ² , ³ . There is growing evidence for an interplay between diabetes and HF. The Framingham Heart Study first demonstrated that men and women with diabetes had a two‐ and five‐fold increased risk of incident HF, respectively, independent of age, hypertension, dyslipidemia, or prior coronary disease ⁴ . In the Reduction of Atherothrombosis for Continued Health Registry, diabetes adversely affects the risk of hospitalization for HF (HHF) in patients at high risk of atherothrombosis or established atherothrombosis over a mean follow‐up of 4 years ⁵ . Thereafter, a meta‐analysis conducted by Dauriz et al. which analyzed more than 100,000 total deaths among nearly 400,000 individuals with HF showed that the survival of patients with coexistent HF and diabetes is significantly lower than that of a non‐diabetic population ⁶ . Notably, major clinical trials have shown that intensive glucose‐lowering strategies did not translate into better cardiovascular outcomes in patients with diabetes ⁷ , ⁸ , ⁹ , ¹⁰ .

Dysglycemia induced myocardial abnormalities can manifest at a very early stage of diabetes in the absence of atherosclerotic cardiovascular disease (ASCVD); hence, HF may be the initial presentation of cardiovascular disease (CVD) in many patients with diabetes ¹¹ . Both Eastern and Western clinical studies indicate that unfavorable preclinical cardiac remodeling begins with glucose intolerance and insulin resistance and progresses with greater dysglycemia ¹² , ¹³ . Recent Asian cohort analyses, such as EMPA‐REG ASIAN ¹⁴ , and post hoc subgroup analysis from the DELIVER (Dapagliflozin Evaluation to Improve the Lives of Patients with Preserved Ejection Fraction Heart Failure) trial, EMPEROR‐Reduced (Empagliflozin Outcome Trial in Patients with Chronic Heart Failure and a Reduced Ejection Fraction) trial, and the DAPA‐HF (Dapagliflozin and Prevention of Adverse Outcomes in Heart Failure) trial ¹⁵ , ¹⁶ , ¹⁷ , confirm that SGLT2 inhibitors reduce the risk of cardiovascular death or worsening HF events and are well tolerated among participants enrolled in both Asia and other geographic regions. These data provide compelling support for initiating SGLT2 inhibitors in Asian patients with T2D who are at risk for HF. Thus, early identification of individuals with diabetes who are at elevated risk for HF is important; however, it presents significant challenges in clinical settings, particularly among patient populations that do not exhibit clinical signs or symptoms of HF.

In the asymptomatic diabetic population without overt heart disease, the prevalence of diastolic and systolic dysfunction can reach 47% and 32%, respectively, as reported in a tissue Doppler imaging‐based cohort among a selected population with normal ejection fractions and controlled blood pressure ¹⁸ . Echocardiographic measurements are now widely used in current guidelines to define preclinical cardiac dysfunction; however, they do not appear to be cost‐effective as a clinical screening tool. The implementation of available strategies to detect early asymptomatic patients with diabetes and a heightened risk of HF is suboptimal and urgently needed.

RATIONALE OF THE DM‐HEART COHORT STUDY

Patients with diabetes face elevated risks of developing both asymptomatic cardiac abnormalities (stage B heart failure) and symptomatic manifestations (stage C heart failure). The recent American Diabetes Association (ADA) “Standards of Care in Diabetes” guideline recommended screening diabetic adults through natriuretic peptide (NP) measurements or N‐terminal pro‐brain natriuretic peptide (NT‐proBNP) for early detection and prevention of stage C HF ¹⁹ . However, among the currently available diabetes‐related HF risk prediction tools, we observed that NT‐proBNP is not commonly included in risk assessment ²⁰ .

In this prospective, observational trial, we intend to assess the prognostic ability of clinically available parameters integrated with NT‐proBNP for predicting incident HF events in asymptomatic patients with type 2 diabetes (T2D) and HF risks. Furthermore, we aim to determine the optimal NT‐proBNP cut‐off value for predicting incident HF events in Asian patients with diabetes (Figure S1).

METHODS

Study design

Overview

The Diabetes Mellitus Heart Failure Events Analysis and Risk Tracking (DM‐HEART) trial comprises two distinct arms. The prospective arm is a multicenter, observational cohort study conducted at National Taiwan University Hospital (NTUH), Far Eastern Memorial Hospital (FEMH), and MacKay Memorial Hospital (MMH) in Taipei, Taiwan. The retrospective arm utilizes data from MMH as the derivation group and NTUH as the validation group. This arm aims to develop and validate an NT‐proBNP‐based heart failure prognostic model among patients with T2D in preclinical stages (stage A or B), thereby enabling healthcare providers to better identify diabetic patients at risk and optimize treatment strategies. The study protocol was approved by the Institutional Review Board (NTUH IRB No: 2104041RINB; FEMH IRB No:109190‐E; MMH IRB No: 20MMHIS365e) at each site and registered on ClinicalTrials.gov (registration number: NCT05866380). The recruitment process in the three centers was standardized, including the essential parts of the baseline evaluations, specimen collection and laboratory tests to be performed. The study design is summarized in Figure 1. The research reported in the paper adhered to the guidelines for reporting observational studies, the Strengthening the Reporting of Observational Studies in Epidemiology Statement.

Figure 1.

Derivation of the DM‐HEART cohort study. HF, heart failure; NT‐proBNP, N‐terminal pro‐brain natriuretic peptide.

Study population

The DM‐HEART study enrolled patients with T2D from outpatient clinics between January 2021 and June 2024, with a minimum follow‐up period of 12 months. Men and women aged 20 years and above residing at the DM‐HEART cohort sites will be invited to participate in the study. This age group was chosen for three reasons: first, the prevalence of young‐onset diabetes has been on the rise; second, current healthcare interventions contribute to the extended lifespan of individuals living with T2D; and third, including patients from a broader range of age groups contributes to a comprehensive understanding of the study outcomes. The diagnosis of T2D will be verified by the use of anti‐diabetic drug or insulin, physician report, or blood glucose levels.

Sample size justification

The prevalence of CVD among individuals with diabetes is from 3% to 25% and the prevalence of HF is from 0.5% to 6% based on age in a Taiwan Nationwide Survey ²¹ . In the WATCH‐DM study, patients with T2D free from baseline HF presented with an average of 3.6% HF event rate. With higher risk for cardiovascular events in Asian patients with hypertension when compared with their Western counterparts, we expected slightly higher clinical event rates for HF in our high‐risk Asian T2D cohort, with an estimated clinical event rate ranging from 4% to 5% annually. By using the clinical predictors proposed in the WATCH‐DM study, along with NT‐proBNP, we estimated an overall minimum of 100 events to reach statistical significance, which may result in approximately 2000 participants in any single dataset in our study design.

Definition of high risk

High risk was defined as the presence of T2D and any one of the following additional risk factors:

• Age ≥ 50 years

• Body mass index (BMI) ≥ 25 kg/m²

• Estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73 m²

• Albuminuria ≥ 30 mg/day

• Hyperlipidemia (low‐density lipoprotein [LDL] > 130 mg/dL) or on lipid‐lowering therapy

• Hypertension (blood pressure > 140/90 mmHg at enrolment) or on antihypertensive therapy

• T2D diagnosed over 10 years.

Eligibility

At baseline, patients diagnosed with T2D for more than 6 months will be screened for inclusion in the study. The aim of the DM‐HEART study is to include a broad range of patients with a high risk of HF in the context of T2D, including those who either have a history of CVDs including prior myocardial infarction (MI), prior percutaneous coronary intervention (PCI), positive stress test, coronary stenosis more than 50% by invasive/computed tomography coronary angiography, prior coronary artery bypass grafting (CABG), or those who exhibit risk factors associated with HF. The aforementioned risk factors for HF are defined based on the WATCH‐DM Score published by Segar et al. ²² which include: advanced age, abnormal lipid profile, elevated blood pressure, overweight/obesity, elevated blood glucose levels, and prolonged QRS duration. In the screening phase, patients with T2D and any one of the previously defined clinical conditions will be defined as the HF high‐risk group and can be enrolled. The inclusion and exclusion criteria are detailed in Table 1.

Table 1.

Inclusion and exclusion criteria of DM‐HEART

Inclusion criteria

Clinical diagnosis of type 2 diabetes (for at least 6 months) Age ≥ 20 years Free at baseline of diagnosed HF Patients who had either established cardiovascular disease or additional cardiovascular risk factors, including age ≥ 50 years, BMI ≥25 kg/m2, eGFR <60 mL/min/1.73m2, albuminuria ≥30 mg/day, hyperlipidemia (defined as LDL‐C >130 mg/dL or being on lipid‐lowering therapy), hypertension (defined as BP >140/90 mmHg at enrollment or being on antihypertensive therapy), and T2D diagnosed more than 10 years prior. Baseline patient‐level clinical data available Willingness to provide informed consent

Exclusion criteria
Known HF hospitalization history Known impaired LV systolic function (LVEF < 40%) by any imaging modality Any terminal disease other than diabetes lowering the patient's life expectancy to less than 2 years eGFR <30 mL/min/1.73 m2 Pregnant women

BMI, body mass index; BP, blood pressure; eGFR, estimated glomerular filtration rate; LDL‐C, low‐density lipoprotein; LV, left ventricle; LVEF, left ventricular ejection fraction.

Informed consent

The approval of the study protocol from the local IRB and consent forms was obtained from each site. Patients were identified at the trial center, and all patients provided written informed consent before undergoing screening procedures.

Predictors of interest

The WATCH‐DM model included patients with T2D who had no history of HF to develop a machine learning‐derived risk score for predicting incident HF from a Western population. Building upon the 10 predictors identified in the WATCH‐DM risk score, we will compile a dataset of 34 clinically relevant variables as predictors for the DM‐HEART model.

The candidate predictor variables encompass a range of demographic, clinical, and biochemical factors, including:

• Demographic Variables: Age, sex

• Clinical Measurements: BMI, systolic blood pressure (SBP), diastolic blood pressure, heart rate

• Lifestyle Factors: Current smoking status

• Metabolic Parameters: Fasting glucose, glycated hemoglobin (HbA1c), total cholesterol, LDL, high‐density lipoprotein (HDL), triglycerides

• Liver and Kidney Function: Alanine aminotransferase, aspartate transaminase, creatinine levels, eGFR, urine albumin‐to‐creatinine ratio (UACR)

• Cardiac Biomarkers: NT‐proBNP

• Cardiovascular History: Previous MI, prior PCI, prior CABG, established peripheral arterial occlusive disease, previous ischemic stroke, history of atrial fibrillation, history of obstructive sleep apnea

• Comorbid Conditions: Dyslipidemia, hypertension

• Electrocardiographic Findings: Left ventricular hypertrophy (LVH) was assessed using the Sokolow‐Lyon criteria, specifically by evaluating the sum of the S wave in V1 and the R wave in V5 or V6. A total voltage of ≥35 mm indicates LVH. Additionally, we considered electrocardiographic parameters such as P wave duration, PR interval, QRS duration, T wave morphology, QT interval, and corrected QT interval.

Baseline evaluations

All participants will be evaluated at baseline for demographic and medical characteristics (Table 2). The face‐to‐face interview is designed to be comprehensive. Each participant will be asked whether they are willing to self‐administer the Taiwanese version of the World Health Organization Quality of Life questionnaire ²³ ; if they agree, an online questionnaire will be provided.

Table 2.

Core data collection on DM‐HEART cohort study at baseline

Demographic information Age (years) Sex (male/female) Physical examination Waist/Hip circumference (cm) Height, weight and BMI (kg/m2) Systolic blood pressure (mmHg) Diastolic blood pressure (mmHg) Heart rate (bpm) Medical history Duration of diabetes (years) Anti‐diabetic medications Hypertension Dyslipidemia Diabetic complications Cardiovascular comorbidities Behavior factors Cigarettes smoking Alcohol consumption Betel nuts chewing

Laboratory values Fasting blood glucose (mg/dL) HbA1C (%) Total cholesterol (mg/dL) Fasting triglycerides (mg/dL) LDL‐C (mg/dL) HDL‐C (mg/dL) Creatinine (mg/dL) eGFR (mL/min/1.73 m2) NTproBNP (pg/mL) Urine microalbumin (mg/dL) Urine creatinine (mg/dL) Microalbumin/creatinine ratio Electrocardiogram QRS duration (ms)

BMI, body mass index; eGFR, estimated glomerular filtration rate; GPT, glutamic pyruvic transaminase; HbA_1C, glycated hemoglobin A_1C; HDL‐C, high‐density lipoprotein; LDL‐C, low‐density lipoprotein; NTproBNP, N‐terminal pro B‐type natriuretic peptide.

Participant characteristics

A total of 1,755 subjects were enrolled from NTUH (n = 514), FEMH (n = 352), and MMH (n = 889). The cohorts were predominantly male across all sites. Demographic characteristics of the participants, stratified by recruitment site, are summarized in Table 3.

Table 3.

Demographic characteristics of the study participants stratified by recruitment site

	NTUH (n = 514)	FEMH (n = 352)	MMH (n = 889)	Total (N = 1,755)
Sex, n (%)
Male	266 (51.8%)	215 (61.1%)	465 (52.2%)	946 (53.9%)
Female	248 (48.2%)	137 (38.9%)	424 (47.8%)	809 (46.1%)
Age, years
Mean ± SD	63.4 ± 10.3	66.5 ± 9.7	63.1 ± 9.8	–
Range	57–71	25–96	28–88	–

FEMH, Far Eastern Memorial Hospital; MMH, MacKay Memorial Hospital; NTUH, National Taiwan University Hospital.

Outcome definitions

Primary outcome

The primary study endpoint is incident HF, which included new‐onset HF requiring hospitalization or an urgent care visit. The definition of HF event requires the following concurrent criteria: (1) primary admission diagnosis of HF with typical symptoms/signs plus objective congestion (e.g., chest‐X‐ray oedema); (2) NT‐proBNP above age‐stratified cut‐points; and (3) augmentation of oral diuretic therapy, or requirement for intravenous loop diuretics therapy.

Secondary outcomes

The secondary endpoints include the assessment of 3‐point major adverse cardiovascular events, defined as cardiovascular death, non‐fatal myocardial infarction, and non‐fatal stroke. Cardiovascular death is defined as any death resulting from stroke, MI, or other cardiovascular causes (e.g., deaths caused by pulmonary embolism; any sudden death, unless proven to have a non‐cardiovascular cause; death after a vascular operation or procedure; death caused by HF). Any MI or stroke followed by a death in the next 28 days (regardless of the cause) will be considered a fatal MI or stroke. Disease diagnoses and procedures were documented using the International Classification of Diseases, 10th Revision‐Clinical Modification (ICD‐10‐CM; Table S1).

Statistical analysis

Data for continuous variables will be expressed as means ± SD and categorical variables as frequencies and proportions of occurrence. Differences in baseline demographics and anthropometric and cardiometabolic parameters among the risk groups (classified as low [≤9], intermediate [10], and high risk [≥11] groups according to WATCH‐DM study risk scores) will be tested using analysis of variance, with categorical data analyzed using the χ² test or Fisher's exact tests as appropriate. Post hoc comparisons between each group will be performed using Scheffés multiple‐comparison test. The relationship among WATCH‐DM scores, dysglycemic indices (i.e., HbA1c or fasting glucose level), echo‐derived parameters, and biomarkers will be determined using Pearson's correlation analysis, with multi‐variate analysis conducted with appropriate adjustment.

Receiver operating characteristic (ROC) curves will be generated to determine optimal cut‐offs of WATCH‐DM scores, NT‐proBNP levels, and other clinical predictors for predicting various clinical endpoints separately and as a composite, with time‐dependent Harrell's concordance index calculated for each event, respectively. Survival analysis will be performed using Cox regression models, with absolute and adjusted incidence rates presented (IR = events/per 1,000 person‐years). The independent prognostic value between various predictors (such as WATCH‐DM strata, NT‐proBNP cut‐off, or other clinical predictors) and clinical endpoints will be examined by multi‐variate Cox models with backwards elimination (P > 0.10). The hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) for each predictor tested will be reported. The Kaplan–Meier estimate of event free (free from individual or composite clinical endpoints) survival will then be calculated and graphed according to WATCH‐DM or NT‐proBNP strata. We anticipate that missing data will be minimal; therefore, complete case analyses will be performed, including only observations with complete information for all covariates in the models, without applying any imputation procedures. We will further test whether circulating NT‐proBNP levels may improve reclassification for clinical endpoints by comparing the relative clinical importance of outcome‐driven NT‐proBNP cut‐off levels with clinical or WATCH‐DM scoring in determining outcomes.

DISCUSSION

Early identification of HF risk in patients with diabetes is crucial, as subclinical cardiac dysfunction often precedes overt symptoms ²⁴ . B‐type natriuretic peptide (BNP), released from the cardiac ventricles in response to ventricular stretch and hemodynamic stimuli, has been shown to effectively screen diabetic patients for the presence or absence of left ventricular (LV) dysfunction ²⁵ . Previous randomized controlled studies underscore the importance of early NP–guided interventions in high‐risk populations to prevent HF. The PONTIAC (NT‐proBNP Selected PreventiOn of cardiac eveNts in a populaTion of diabetic patients without A history of Cardiac disease) trial assessed neurohumoral therapy in high‐risk diabetic patients with elevated NT‐proBNP levels (>125 pg/mL) but no existing cardiac disease ²⁶ . Over 2 years, the intensified treatment group showed significant reductions in hospitalizations and deaths from cardiac disease, along with improvements in blood pressure and heart rate, indicating the effectiveness of up‐titrating renin–angiotensin system antagonists and beta‐blockers in preventing cardiac events. The STOP‐HF (St. Vincent's Screening TO Prevent Heart Failure) trial assessed the efficacy of a BNP–guided screening program in reducing the incidence of new‐onset HF and prevalence of LV dysfunction among participants with cardiovascular risk factors ²⁷ . Over a mean follow‐up of 4.2 years, intervention‐group participants with BNP levels ≥50 pg/mL underwent echocardiography and received collaborative care, resulting in a reduction in the composite outcome of left ventricular systolic dysfunction, diastolic dysfunction, and incident HF. Unlike PONTIAC and STOP‐HF, which predominantly enrolled Western populations, DM‐HEART focuses on an East Asian cohort with T2D without elevated natriuretic peptide levels. DM‐HEART is a prospective, multicenter, longitudinal cohort study designed to ascertain HF incidence and evaluate the prognostic utility of clinical parameters, including NT‐proBNP, for predicting incident HF among asymptomatic patients with preclinical HF. Additionally, the study aims to provide the optimal NT‐proBNP cut‐off for predicting incident HF in East Asian patients with T2D.

Recent research highlights that NP cut‐off levels for identifying cardiac anomalies in asymptomatic Asian populations are notably lower than those recommended by international guidelines ²⁸ , ²⁹ . This suggests a need for region‐specific diagnostic thresholds to enhance early detection and management of cardiac conditions in these populations. Establishing Asian‐specific NT‐proBNP cut‐offs carries important clinical implications. Population‐tailored thresholds could enhance diagnosis for acute and chronic HF, optimize hospitalization decisions, facilitate early risk stratification, and improve outpatient surveillance, while reducing the risk of underestimation and unnecessary testing—thereby conserving healthcare resources and improving overall outcomes. Asian‐specific NT‐proBNP cut‐offs address population‐specific physiological differences and can inform the refinement of clinical guidelines. For instance, ADA guidelines could integrate ethnicity‐specific screening thresholds and retesting intervals within cardio‐renal risk sections, aligning initiation of therapies such as SGLT2 inhibitors and angiotensin receptor–neprilysin inhibitors with biomarker‐driven criteria. Similarly, European Society of Cardiology guidelines could incorporate ethnicity‐calibrated cut‐offs into diagnostic algorithms for acute and chronic heart failure, HFpEF pathways, and primary prevention strategies. Both organizations should advocate prospective, multinational Asian datasets to refine decision curves, define implementation strategies, and include ethnicity‐referenced ranges in laboratory reports.

Common diabetes comorbidities, including chronic hyperglycemia, hypertension, dyslipidemia, and inflammation, synergistically contribute to diabetic cardiomyopathy and HF by promoting myocardial fibrosis, endothelial dysfunction, and oxidative stress ³⁰ . Asian patients with diabetes exhibit distinct cardiovascular risk profiles compared to Western populations. The ASIAN‐HF study, a multinational registry of symptomatic Asian HF patients, revealed that those with reduced ejection fraction were younger (median age 60 years), leaner (BMI 24.9 ± 5.1 kg/m²), and exhibited a higher prevalence of comorbidities such as hypertension, cardiovascular disease, and diabetes compared to EuroHeart Failure Survey 2 patients. Asian individuals, despite similar BMI and waist circumference, exhibit poorer β‐cell function and greater insulin resistance than White individuals and face higher risks of advanced cardiac remodeling with equivalent clinical risk factors ³¹ , ³² . Echouffo‐Tcheugui et al. systematically reviewed 28 HF risk prediction models, most developed in the general population, and found that the most common predictors included in the final HF prediction models were age, sex, BMI, SBP, blood glucose levels, creatinine, serum albumin, LV hypertrophy (LVH), and coronary artery disease; moreover, the application of novel blood and imaging markers, like NT‐proBNP, interleukin‐6, and a composite of LVH and LV dysfunction, can modestly improve risk prediction ³³ . More recently, Kostopoulos et al. analyzed 43 models developed in patients with T2D for HF prediction and concluded that the top 10 strongest risk factors for incident HF in the diabetes cohort are age, HbA1c level, SBP, sex, BMI, CVD, hypertension, smoking, eGFR, and ethnicity, ranked in order ³⁴ . Notably, atrial fibrillation (AF) is a strong risk factor for HF, and they frequently occur together or complicate each other. In permanent AF, the co‐prevalence of HF accounts for more than 50% of the cases ³⁵ . AF and HF share common pathophysiological features: increased end‐diastolic LV pressure in HF adversely affects atrial hemodynamics, which may lead to AF; further, irregular atrioventricular conduction, in particular high heart rates, can induce LV remodeling and, eventually, HF ³⁶ . Though strongly correlated, the presence of AF does not seem to be considered in existing clinical HF prognostic models. Identifying optimal risk factors for HF prognostic models is essential for effective risk stratification and timely intervention, particularly in asymptomatic patients with type 2 diabetes. However, specific models tailored for Asian diabetes patients are currently lacking.

Several clinical models demonstrate promising performance and applicability in patients with T2D, including the Risk Equations for Complications of T2D (RECODe) ³⁷ , the Thrombolysis in Myocardial Infarction [TIMI] Risk Score for Heart Failure in Diabetes (TRS‐ HFDM) ³⁸ and WATCH‐DM ²² . RECODe incorporates 16 variables, including demographics, medical history, treatment drugs, and biomarkers, to predict the 10‐year risk of micro‐ and macrovascular events, including HF ³⁷ . RECODe was validated in two large cohort studies, the Multi‐Ethnic Study of Atherosclerosis (MESA) and the Jackson Heart Study (JHS), demonstrating similar discriminative ability across different settings ³⁹ . Later, researchers developed the TRS‐HFDM and WATCH‐DM, two integer‐based scores, to estimate incident HF risk in participants with T2D who did not have baseline HF ²² , ³⁸ . The TRS‐HFDM includes only five variable risk scores and has demonstrated good performance to predict short term risk of HF among adults with T2D. Furthermore, the addition of NT‐proBNP levels to the TRS‐ HFDM improved the reclassification of patients with T2D at very high risk of CVD ⁴⁰ . The WATCH‐DM is a machine learning derived model predicting the 5‐year risk of HHF and HF related death, including 10 predictors ²² . Among patients with T2D with no history of HF, this integer‐based WATCH‐DM risk score demonstrated good stratification, with the cumulative incidence of HF increasing from 1.1% in the lowest quintile (total score ≤7) to 17.4% in the highest quintile (total score ≥ 14). Furthermore, patients with diabetes with low‐to‐intermediate risk were better reclassified when WATCH‐DM variables and NT‐proBNP levels were combined ⁴¹ . According to a meta‐analysis using C‐statistics, the above three clinical models all performed well, and RECODe appears to be the most suitable tool for simultaneous risk assessment in patients with T2D at the moment ²⁰ . However, if the prediction model incorporates more clinical factors, the ease of clinical appraisal may decrease. Compared to RECODe, which incorporates 16 variables, the integer‐based WATCH‐DM risk score is simpler, utilizing only 10 predictors. Its robust performance in both risk stratification and prediction led to its application as a stratification tool in the DM‐HEART trial.

Limitations

To aid in the interpretation of the data, it is important to acknowledge the limitations of the current study. First, the study population primarily consists of Taiwanese patients with T2D from three medical centers in Taipei, potentially limiting the generalizability of the data due to the heterogeneity of diabetic phenotypes in different regions across Asia. Second, a single NT‐proBNP measurement may limit dynamic risk assessment in heart failure management. To enhance prognostic capabilities, future studies should focus on serial NT‐proBNP measurements, allowing for more accurate tracking of cardiac function over time. Furthermore, while ECG‐based assessment of LVH provides a rapid evaluation, it may under‐detect structural abnormalities when compared to echocardiography. By highlighting the differences in diagnostic capabilities between these modalities, we aim to provide a clearer context for our findings and emphasize the value of comprehensive assessment strategies in clinical practice.

CONCLUSIONS

Recognition of asymptomatic patients with T2D who had stage A or B HF is important to implement timely and appropriate HF prevention strategies. Current HF prediction models have acceptable discrimination ability in derivation data sets; however, most models were developed based on Western populations. The DM‐HEART study is groundbreaking as the first prospective, large‐scale, multicenter, massive HF screening project involving multiple centers, enabling comprehensive data collection across diverse asymptomatic Asian diabetic populations. This study aims to construct a useful clinical model to help physicians more accurately identify which patients with T2D may be at risk of HF and need early intervention. Given that the prevalence of diabetes has risen globally and is expected to increase further, the burden of HF on the health care system will continue to rise. Early diagnosis, effective management, and targeted interventions can mitigate HF progression and enhance overall well‐being in the asymptomatic diabetic population.

DISCLOSURE

The authors declare no conflict of interest.

Approval of the research protocol: The study protocol was approved by the Institutional Review Board (Mackay Memorial Hospital, IRB No: 20MMHIS365e).

Informed Consent: The authors confirm that patient consent forms have been obtained for this article.

Approval date of Registry and the Registration No. of the study/trial: Approval date of Registry: 2023‐05‐19.

Registration No. of the study/trial: ClinicalTrials.gov (registration number: NCT05866380).

Animal Studies: N/A.

Supporting information

Table S1. Outcome codes in the International Classification with Diseases, 10th edition.

Figure S1. The integrated referral program for patients with diabetes and risk of heart failure. The integrated referral program is important for diabetic patients at high risk for heart failure. This program focuses on identifying associated symptoms and complex risk factors to facilitate the early detection of high‐risk individuals and ensure seamless referrals. By integrating primary care, specialty medical services, and community resources, the program can improve patient outcomes, reduce the risk of complications, and enhance quality of life.

DATA AVAILABILITY STATEMENT

The data are not publicly available due to privacy concerns. For any inquiries regarding the data, please contact corresponding authors upon reasonable request.