Predictive value of echocardiography in Type 2 diabetes

Aims Echocardiography is suggested in the diagnostic work-up of patients with Type 2 diabetes (T2D). We investigated which echocardiographic parameters that best predicted cardiovascular disease (CVD) and whether this was persistent in both genders in a large cohort of outpatients with T2D. Methods and results We performed comprehensive echocardiography in 933 patients with T2D followed at specialized out-patients clinics in Copenhagen, Denmark. Follow-up was performed using national registries and included admission with future CVD events and non-CVD death as competing risk. Median follow-up was 4.8 years and 138 CVD events occurred. In univariable and multivariable analyses, a wide range of structural, diastolic, and systolic measurements predicted CVD including mean E/e' [hazard ratio (HR) 1.06, 95% confidence interval: (1.03-1.10), P < 0.001, C-statistics 0.74 (0.70-0.78)] and global longitudinal strain (GLS) [1.10 (1.01-1.20), P = 0.03, C-statistics 0.73 (0.69-0.77)]. However, this was modified by gender. In men, mean E/e' remained the strongest predictor in multivariable analyses and performed best measured by highest C-statistics [HR 1.15, 95% confidence interval: (1.08-1.21), P < 0.001, C-statistics 0.75 (0.71-0.80)] whereas in women this was GLS [1.39 (1.14-1.70), P = 0.001, C-statistics 0.79 (0.70-0.87)]. These findings persisted when excluding patients with known heart disease and when regarding all-cause mortality as a competing risk. Conclusion A range of echocardiographic parameters predicted CVD in patients with Type 2 diabetes, however, in multivariable analyses, mean E/e' was the strongest predictor and had the highest model performance. Importantly, this study identifies a hitherto undescribed gender interaction as mean E/e' performed best in men, whereas in women this was GLS.