A total of 303 patients were enrolled, comprising 229 HF patients and 73 HC individuals. Among the HF patients, 98 HFrEF patients and 91 HFpEF patients were included based on the classification criteria of the 2021 European Society of Cardiology (ESC) HF guidelines. Table 1 shows the clinical characteristics of HF patients and HC individuals. Compared with the HC group, the HF patients were older (P < 0.001), more likely to be men (P < 0.001), and had a higher level of WBC (P = 0.002) and NLR (P < 0.001) but had lower albumin (Alb) levels (P < 0.001). More importantly, SI was significantly different between the HF and HC groups (Figure S1A).
Table 1
Clinical characteristics of the HF and HC groups.
| HF N = 229 | HC N = 73 | p-Value |
Age (years) | 68.4 ± 12.3 | 53.5 ± 11.8 | <0.001 |
Male (%) | 157 (68.6) | 33 (45.2) | <0.001 |
BMI (kg/m2) | 24.7 ± 4.5 | 24.2 ± 3.2 | 0.603 |
SBP (mmHg) | 135 ± 75 | 128.5 ± 16.1 | 0.855 |
DBP (mmHg) | 79 ± 15 | 78.0 ± 11.2 | 0.702 |
Smoking (%) | 70 (30.6) | 19 (26.0) | 0.459 |
Alcohol (%) | 39 (17.4) | 7 (9.6) | 0.123 |
ALT (U/L) | 35.1 ± 70.4 | 20.3 ± 11.1 | 0.192 |
AST (U/L) | 36.8 ± 82.4 | 18.9 ± 5.5 | <0.001 |
LDH (U/L) | 234.7 ± 167.7 | 173.2 ± 32.7 | <0.001 |
STB (umol/L) | 17.3 ± 9.3 | 10.5 ± 4.5 | <0.001 |
Alb (g/L) | 36.0 ± 4.4 | 49.8 ± 3.3 | <0.001 |
Glb (g/L) | 26.6 ± 5.1 | 27.6 ± 4.3 | 0.100 |
TC (mmol/L) | 4.1 ± 1.1 | 4.7 ± 0.9 | <0.001 |
TG (mmol/L) | 1.4 ± 1.0 | 2.3 ± 2.3 | <0.001 |
HDL (mmol/L) | 1.0 ± 0.3 | 1.2 ± 0.3 | <0.001 |
LDL (mmol/L) | 2.4 ± 0.9 | 2.5 ± 0.7 | 0.022 |
Apo-A1 (g/L) | 1.1 ± 0.6 | 1.2 ± 0.2 | <0.001 |
Apo-B (g/L) | 0.9 ± 0.6 | 0.9 ± 0.2 | 0.003 |
BUN (mmol/L) | 7.8 ± 3.7 | 4.8 ± 1.3 | <0.001 |
SI (%) | 67.9 ± 13.0 | 98.6 ± 31.5 | <0.001 |
Hb (g/L) | 132.4 ± 22.8 | 135.2 ± 21.6 | 0.210 |
WBC (10^9/L) | 6.8 ± 2.3 | 5.8 ± 1.5 | 0.002 |
NLR | 3.9 ± 4.2 | 1.7 ± 0.8 | <0.001 |
RDW (%CV) | 14.0 ± 1.8 | 13.1 ± 2.2 | <0.001 |
Abbreviations: |
Alb, albumin; ALT, alanine aminotransferase; Apo-A1, apolipoprotein-A1; Apo-B, apolipoprotein-B; AST, aminotransferase; BMI, body mass index; BUN, blood urea nitrogen; DBP, diastolic blood pressure; Glb, globulin; Hb, haemoglobin; HC, healthy control; HDL, high-density lipoprotein; HF, heart failure; LDH, lactate dehydrogenase; LDL, low-density lipoprotein; NLR, neutrophil-to-lymphocyte ratio; RDW, red blood cell distribution width; SBP, systolic blood pressure; SI, sarcopenia index; STB, serum total bilirubin; TC, total cholesterol; TG, triacylglycerol; WBC, white blood counts. |
Table 2 shows the clinical characteristics of the HFrEF and HFpEF groups. The individuals in the HFpEF group were older (P < 0.001), more likely to be women (P < 0.001), and had a higher prevalence of hypertension (P = 0.007) and AF (P < 0.001) but had lower levels of NT-proBNP (P < 0.001), BUN (P = 0.012), Hb (P < 0.001), and QRS interval (P < 0.001). Importantly, SI was significantly lower in the HFpEF group than in the HFrEF group (Figure S1B).
Table 2
Clinical characteristics of HFrEF and HFpEF groups.
| HFrEF N = 98 | HFpEF N = 91 | p-Value |
Age (years) | 63.6 ± 13.9 | 71.7 ± 11.4 | <0.001 |
Male (%) | 77 (78.6) | 46 (50.5) | <0.001 |
BMI (kg/m2) | 24.4 ± 5.1 | 24.9 ± 4.4 | 0.182 |
SBP (mmHg) | 124.2 ± 18.8 | 134.9 ± 21.9 | 0.001 |
DBP (mmHg) | 80.5 ± 13.7 | 79.0 ± 14.2 | 0.873 |
cTnI (pg/ml) | 0.11 ± 0.31 | 0.06 ± 0.15 | 0.015 |
NT-proBNP (pg/ml) | 7204.3 ± 7995.9 | 3306.4 ± 3143.4 | <0.001 |
Smoking (%) | 40 (40.8) | 15 (16.5) | <0.001 |
Alcohol (%) | 24 (24.5) | 10 (11.0) | 0.016 |
Hypertension (%) | 49 (50.0) | 63 (69.2) | 0.007 |
CADs (%) | 50 (51.0) | 40 (44.0) | 0.202 |
AF (%) | 20 (20.4) | 61 (67.0) | <0.001 |
ALT (U/L) | 50.6 ± 112.4 | 24.7 ± 20.3 | 0.210 |
AST (U/L) | 50.4 ± 132.2 | 26.3 ± 11.9 | 0.544 |
STB (umol/L) | 19.6 ± 10.9 | 15.9 ± 9.1 | 0.052 |
Alb (g/L) | 36.5 ± 4.4 | 35.2 ± 4.7 | 0.174 |
Glb (g/L) | 26.1 ± 4.8 | 27.8 ± 5.9 | 0.111 |
BUN (mmol/L) | 8.3 ± 4.1 | 7.0 ± 3.2 | 0.012 |
SI (%) | 71.8 ± 13.7 | 62.3 ± 12.0 | <0.001 |
eGFR | 76.5 ± 21.8 | 73.1 ± 21.4 | 0.795 |
Hb (g/L) | 138.6 ± 25.6 | 127.0 ± 19.3 | <0.001 |
WBC (10^9/L) | 6.7 ± 1.8 | 6.7 ± 2.5 | 0.251 |
NLR | 3.5 ± 2.9 | 3.8 ± 2.9 | 0.234 |
QRS interval (ms) | 123.3 ± 27.7 | 108.0 ± 22.5 | <0.001 |
Univariate logistic regression analysis showed that lower SI was significantly associated with heart failure at hospital admission (OR 0.924, 95% CI 0.904–0.944, P < 0.001; Table S2). Variables, such as gender, alanine aminotransferase (ALT), aminotransferase (AST), lactate dehydrogenase (LDH), serum total bilirubin, Alb, total cholesterol, triacylglycerol (TG), high-density lipoprotein (HDL), BUN, SI, WBC, NLR, and RDW, were included in univariate logistic regression. Multivariate logistic regression showed that lower levels of SI was still an independent risk factor for HF (OR 0.921, 95% CI 0.879–0.965, P = 0.001).
Among the 229 HF patients, SI had a significant negative correlation with LVEF (r = -0.256, P < 0.001, Figure S2). Moreover, SI was investigated in the HFrEF and HFpEF groups. The univariate logistic regression analysis showed that lower SI was significantly associated with an increased prevalence of HFpEF (OR 0.947, 95% CI 0.923–0.972, P < 0.001; Table 3). Variables, such as age, gender, SBP, NT-proBNP (lnNT-proBNP), smoking, alcohol, hypertension, AF, BUN, SI, Hb, and QRS interval, were included in univariate logistic regression. Multivariate logistic regression showed that a low level of SI was still an independent risk factor for HFpEF (OR 0.948, 95% CI 0.914–0.983, P = 0.004).
Table 3
Univariate and multivariate logistic regression of factors associated with HFpEF.
Variables | Univariate logistic regression | Multivariate logistic regression |
OR (95%CI) | p-Value | OR (95%CI) | p-Value |
Age | 1.054 (1.026–1.082) | <0.001 | 1.026 (0.987–1.068) | 0.199 |
Male | 0.279 (0.148–0.525) | <0.001 | 0.612 (0.226–1.654) | 0.333 |
BMI | 1.029 (0.967–1.095) | 0.365 | | |
SBP | 1.026 (1.011–1.041) | 0.001 | 1.027 (1.004–1.051) | 0.022 |
DBP | 0.997 (0.997–1.018) | 0.784 | | |
cTnI | 0.442 (0.089–2.184) | 0.316 | | |
LnNT-proBNP | 0.294 (0.150–0.577) | <0.001 | 0.136 (0.046–0.401) | <0.001 |
Smoking | 0.286 (0.144–0.568) | <0.001 | 0.479 (0.159–1.449) | 0.193 |
Alcohol | 0.381 (0.171–0.849) | 0.018 | 0.615 (0.162–2.340) | 0.476 |
Hypertension | 2.250 (1.240–4.038) | 0.008 | 1.537 (0.625–3.780) | 0.349 |
CADs | 0.688 (0.387–1.223) | 0.203 | | |
AF | 7.930 (4.109–15.305) | <0.001 | 6.336 (2.511–15.988) | <0.001 |
ALT | 0.990 (0.979–1.001) | 0.072 | | |
AST | 0.983 (0.966–1.002) | 0.076 | | |
STB | 0.969 (0.939-1.000) | 0.053 | | |
Alb | 0.951 (0.890–1.015) | 0.130 | | |
Glb | 1.053 (0.993–1.116) | 0.086 | | |
BUN | 0.905 (0.828–0.989) | 0.028 | 0.912 (0.795–1.046) | 0.187 |
SI | 0.947 (0.923–0.972) | <0.001 | 0.948 (0.914–0.983) | 0.004 |
eGFR | 0.998 (0.986–1.010) | 0.745 | | |
Hb | 0.977 (0.963–0.991) | 0.001 | 0.980 (0.962–0.999) | 0.043 |
WBC | 0.981 (0.855–1.125) | 0.782 | | |
NLR | 1.062 (0.969–1.163) | 0.198 | | |
QRS interval | 0.975 (0.963–0.988) | <0.001 | 0.980 (0.963–0.997) | 0.020 |
Using LASSO to screen variables for the prediction model (Fig. 1A and 1B) and final nine predictor variables, including age, gender, SBP, smoking, NT-proBNP (lnNT-proBNP), SI, Hb, QRS interval, and AF, were screened to predict the risk of HFpEF compared with HFrEF in hospitalized patients. Nomogram was drawn to show the results based on the predictor variables (Fig. 2). Among the different predictor subjects, the possible risk of HFpEF might be measured by calculating the sum of the total scores corresponding to each different variable in the predictor object.
The area under the curve (AUC) of the ROC curve for the model constructed with LASSO in the original data was 0.902, suggesting that the prediction model was highly discriminatory and had a high predictive power for HFpEF (Fig. 1C). Additionally, the model was validated internally by Bootstrap. The calibrated C-statistic was 0.903 after 500 resamplings, and the calibration curve was distributed approximately along the reference line (Fig. 1D), which indicated a good fit for the prediction model. Altogether, the model had a high level of discrimination and calibration.
To evaluate the impact of comorbidities in the prediction model for HFpEF, two prediction models were compared for their prediction effectiveness. Age, gender, SBP, smoking, NT-proBNP (lnNT-proBNP), SI, Hb, QRS interval, and AF were included In Model 1while Model 2 included age, gender, SBP, smoking, NT-proBNP (lnNT-proBNP), Hb and QRS interval. The ROC curve analysis was used to compare the prediction effectiveness of the two models, and the results indicated that the prediction effectiveness of Model 1 was significantly better than that of Model 2 (AUC, 0.902 vs. 0.855, P < 0.01; Fig. 3A). Likewise, the net benefit of all models surpassed the reference model throughout the threshold ranges, and Model 1 outperformed Model 2 by using DCA (Fig. 3B).