Baseline characteristics of the study population
A total of 385 T2DM patients with HFrEF were enrolled in this study. After 6-12 months, 167 (43.4% [95% CI 38.4%~48.3%]) patients developed HFrecEF and another 218 (56.6% [95% CI 51.7%~61.6%]) patients remained HFrEF. Compared with persistent HFrEF patients, HFrecEF patients tended to be younger and less frequently have an ischemic etiology. They also appeared to have higher blood pressure, total and LDL cholesterol levels. Other clinical characteristics including sex, history of hypertension, myocardial infarction, atrial fibrillation, smoking habits, New York Heart Association (NYHA) grades, renal function, and medication treatments were similar between the 2 groups (Table 1).
Changes in LV function and geometric
At baseline, T2DM patients with HFrecEF had similar LV function as persistent HFrEF patients (33.03% ± 5.59% vs. 33.83% ± 6.51%, P=0.205), but tended to have smaller LV volumes (EDV and ESV indices, all P<0.001). During the echocardiogram follow-up, HFrecEF patients exhibited pronounced EF recovery (33.03% ± 5.59% to 52.28% ± 7.70%, P<0.001) and reverse LV remodeling (ΔEDV index: -23.94 ± 23.32 mL/m2, ΔESV index: -31.75 ± 22.48 mL/m2, Supplementary Table I).
Association between HbA1c and HFrecEF
T2DM patients with HFrecEF had significantly lower HbA1c level than those with persistent HFrEF (6.4% [IQR 5.8%~7.1%] vs. 6.8% [IQR 6.1%~7.8%], P=0.001), Figure 2A). Lower HbA1c levels in patients with HFrecEF versus HFrEF (Figure 2B) were only observed in the ischemic subgroup (6.4% [IQR 5.8%~7.2%] vs. 6.9% [IQR 6.2%~8.0%], P=0.003) rather than those with non-ischemic etiology (6.4% [IQR 5.8%~7.1%] vs. 6.6% [IQR 6.0%~7.4%], P=0.178).
Univariate analysis (Supplementary Table II) revealed that predictors for HFrecEF in T2DM patients were younger age (OR: 0.758 [95% CI 0.630~0.907]), non-ischemic etiology (OR: 1.755 [95% CI 1.156~2.673]), higher diastolic blood pressure (OR: 1.153 [95% CI 1.005~1.329]), total cholesterol levels (OR: 1.323 [95% CI 1.087~1.620]) and LDL cholesterol levels (OR: 1.466 [95% CI 1.154~1.879]), as well as smaller LV volumes (OR: EDV index, 0.844 [95% CI 0.777~0.912], per 10 mL/m2, ESV index, 0.904 [95% CI 0.827~0.982], per 10 mL/m2). HbA1c levels were inversely associated with HFrecEF both when treated as continuous (OR: 0.814 [95% CI 0.697~0.943]) and categorical variables (OR: 0.407 [95% CI 0.218~0.746], HbA1c > 8% vs. ≤ 6%).
Multivariate analysis (Table 2) showed that age, non-ischemic etiology, LDL cholesterol levels, baseline EDV index and HbA1c levels were associated with the development of HFrecEF. When treated as categorical variables, patients with HbA1c > 8.0% corresponded to a 57.5% (OR: 0.425 [95% CI 0.205~0.863]) decreased likelihood of HFrecEF as compared to those with ≤ 6%. Restricted cubic spline analysis showed consistently decreased incidence of HFrecEF with elevated HbA1c levels peaked at 8.2% (Supplementary Figure I).
Subgroup analyses demonstrated that T2DM patients who were male, younger, with ischemic etiology, smoking habits, history of hypertension, lower BMI and poorer renal function were more likely to be affected by HbA1c than respective counterparts in the development of HFrecEF. The association between HbA1c and HFrecEF was affected by smoking habits (P for interaction = 0.018) and history of hypertension (P for interaction = 0.038, Figure 3).
All the sensitivity analyses are presented in Table 3. After multivariate adjustment, the association between HbA1c levels and HFrecEF persisted after exclusion of patients with previous myocardial infarction (Model 1), receiving coronary revascularization (Model 2) or with ‘partial EF recovery’ (a second EF > 40% but an absolute EF improvement <10%, Model 3). The association remained significant when HFrecEF was alternatively defined by EF recovery to ≥ 50% (Model 4). When entering the mean HbA1c level of all the available tests at baseline and throughout the echocardiogram follow-up instead of the one obtained at admission, it remained inversely associated with the development of HFrecEF (Model 5).
Glycemic control and HF prognosis
The cohort was followed-up for 3.3±1.7 years, with a total of 1261.1 patient-years. Patients with poor glycemic control were more likely to experience CV death or HF rehospitalization (P=0.026, Table 4 and Figure 4A). In fully adjusted models, patients with HbA1c >8% had a 2.8-fold increased risk of the composite end point (HR: 2.778 [95% CI 1.470~5.251], P=0.002) compared to those with HbA1c ≤ 6% (Table 4).
CV death occurred in 17.1% (66 of 385) of the cohort. Age- and sex-adjusted 5-year CV mortality was 10.8%, 19.0%, 19.0% and 34.7% in patients with HbA1c≤6%, 6%~7%, 7~8% and >8%, respectively. There was no significant difference in CV mortality between patients with different glycemic control levels (P=0.097, Figure 4B).
Of note, T2DM with HFrecEF had substantially lower risk of the composite endpoint (HR: 0.341 [95% CI 0.199~0.584], Figure 4C) and CV death (HR: 0.406 [95% CI 0.210~0.785]) compared to those with persistent HFrEF (Figure 4D).