Predicting heart failure in asymptomatic diabetes: derivation and internal validation of a clinical prediction model for early detection of diabetic cardiomyopathy.
To develop and internally validate a clinical prediction model for identifying asymptomatic type 2 diabetes (T2DM) patients at risk of incident heart failure (HF) or progression of subclinical cardiac dysfunction.
This single-center retrospective study included 326 asymptomatic T2DM patients with preserved LVEF, all with ≥3 echocardiographic assessments over 24 months. The primary outcome was a composite of incident clinical HF (classified as HFpEF or HFrEF based on symptoms, hospitalization, natriuretic peptides, and follow-up echo) and imaging-defined progression of subclinical dysfunction. Candidate predictors (clinical variables, GLS, LASr, E/e', LAVI, LVMI, NT-proBNP, galectin-3) were pre-specified. Missing data were handled by multiple imputation. A risk score was derived using multivariable Cox regression with Fine-Gray competing risk analysis, and optimism was assessed via bootstrap, 5-fold cross-validation, and internal split-sample validation.
Over 24 months, LV GLS modestly declined (17.4 ± 2.0% to 16.9 ± 2.1%), while LASr deteriorated more (38.5 ± 7.2% to 34.0 ± 6.5%). Independent predictors of the composite outcome were LASr ≤24% (HR 3.58), NT-proBNP ≥120 pg/mL (HR 2.48), galectin-3 ≥15 ng/mL (HR 1.79), age ≥70 years, diabetes duration ≥12 years, BMI ≥30 kg/m², and UACR ≥60 mg/g; SGLT2i/GLP-1 RA use was associated with lower observed risk. During follow-up, 40 patients reached the composite outcome (18 clinical HF: 14 HFpEF, 4 HFrEF; 22 imaging-only progression). The scoring system stratified patients into low- (≤3 points), intermediate- (4-6), and high-risk (≥7) groups, with 24-month cumulative incidences of 4.2%, 11.7%, and 27.5% (Gray's test P<0.001). The full model had a C-statistic of 0.835 (bootstrap-corrected 0.828) and good calibration (Brier score 0.068). Adding LASr and galectin-3 improved discrimination (ΔC=0.058 and 0.012, both P<0.05).
Integrating LASr, galectin-3, and clinical risk factors may help identify asymptomatic T2DM patients at higher risk of incident HF or subclinical progression. This internally validated model may support preliminary risk stratification for suspected early diabetic cardiomyopathy, but external validation and clinical utility assessment are needed before routine use.
This single-center retrospective study included 326 asymptomatic T2DM patients with preserved LVEF, all with ≥3 echocardiographic assessments over 24 months. The primary outcome was a composite of incident clinical HF (classified as HFpEF or HFrEF based on symptoms, hospitalization, natriuretic peptides, and follow-up echo) and imaging-defined progression of subclinical dysfunction. Candidate predictors (clinical variables, GLS, LASr, E/e', LAVI, LVMI, NT-proBNP, galectin-3) were pre-specified. Missing data were handled by multiple imputation. A risk score was derived using multivariable Cox regression with Fine-Gray competing risk analysis, and optimism was assessed via bootstrap, 5-fold cross-validation, and internal split-sample validation.
Over 24 months, LV GLS modestly declined (17.4 ± 2.0% to 16.9 ± 2.1%), while LASr deteriorated more (38.5 ± 7.2% to 34.0 ± 6.5%). Independent predictors of the composite outcome were LASr ≤24% (HR 3.58), NT-proBNP ≥120 pg/mL (HR 2.48), galectin-3 ≥15 ng/mL (HR 1.79), age ≥70 years, diabetes duration ≥12 years, BMI ≥30 kg/m², and UACR ≥60 mg/g; SGLT2i/GLP-1 RA use was associated with lower observed risk. During follow-up, 40 patients reached the composite outcome (18 clinical HF: 14 HFpEF, 4 HFrEF; 22 imaging-only progression). The scoring system stratified patients into low- (≤3 points), intermediate- (4-6), and high-risk (≥7) groups, with 24-month cumulative incidences of 4.2%, 11.7%, and 27.5% (Gray's test P<0.001). The full model had a C-statistic of 0.835 (bootstrap-corrected 0.828) and good calibration (Brier score 0.068). Adding LASr and galectin-3 improved discrimination (ΔC=0.058 and 0.012, both P<0.05).
Integrating LASr, galectin-3, and clinical risk factors may help identify asymptomatic T2DM patients at higher risk of incident HF or subclinical progression. This internally validated model may support preliminary risk stratification for suspected early diabetic cardiomyopathy, but external validation and clinical utility assessment are needed before routine use.