Abstract
Aims
This study aimed to establish a clinical prediction model for evaluating the risk of heart failure (HF) with preserved ejection fraction (HFpEF) occurrence in patients with acute myocardial infarction (AMI).
Methods
We selected AMI patients with left ventricular ejection fraction ≥ 50% without a history of HF from the MIMIC-III database (training cohort) and those from The First Affiliated Hospital of Jinan University (validation cohort). Eligibility criteria for identifying patients who develop HFpEF were based on ICD-9 diagnosis codes in the training cohort and the 2021 European Society of Cardiology HF Guidelines in the external validation cohort. The LASSO-multivariate logistic regression analysis was performed to identify the predictive factors for establishing the nomogram prediction model. Then, this model was evaluated by discrimination, calibration, clinical effectiveness, and internal validation. Finally, using the validation cohort, the model was externally validated and updated.
Results
201 patients were included in the training cohort and 65 in the validation cohort. Older age, higher heart rate, higher neutrophil-to-lymphocyte ratio, and lower total bilirubin were identified as independent predictors. The area under the receiver operator characteristic curve (AUC) of the model and the bootstrap validation were 0.823 and 0.804, respectively. Both the prediction model’s and the bootstrap validation’s calibration curve were in good agreement with the standard reference line, and the Hosmer-Lemeshow test showed P = 0.183. The decision curve analysis showed that using this model could produce a net benefit. In external validation, the AUC was 0.749, with a calibration slope of 0.833, a calibration intercept of 3.746, and a Brier score of 0.572. After being updated, the external calibration performance has improved significantly (slope 1, intercept 0, Brier score 0.185).
Conclusion
Our prediction model provides an additional perspective into the early identification of high-risk patients for HFpEF occurrence following AMI.
Keywords: Acute myocardial infarction, Heart failure, Preserved ejection fraction, Prediction model
1. Introduction
The various complications after acute myocardial infarction (AMI) significantly affect the long-term prognosis and survival quality of patients. For example, the heart of AMI patients would undergo a series of dynamic pathological remodeling, resulting in the gradual development of corresponding structural and/or functional abnormality, and ultimately in heart failure (HF).[1].
In recent decades, there have been significant changes in the epidemiological characteristics of HF for AMI patients, with one crucial manifestation being a marked increase in the proportion of HF with preserved ejection fraction (HFpEF), accompanied by a decline in HF with reduced ejection fraction (HFrEF) and HF with mid‐range ejection fraction (HFmrEF). A large retrospective cohort study of 199,851 patients showed that the incidence of HF in AMI patients decreased from 46% (1996–1997) to 28% (2008). However, the proportion of HFpEF as a percentage of total HF types increased from 18% to 30%, whereas the remaining HF subtypes (HFrEF and HFmrEF) decreased from 82% to 70%. [2] A more recent retrospective cohort study that specifically evaluated HF subtypes among AMI patients (2007–2016) reported that HFpEF accounted for 41% of cases, compared with 32% for HFrEF and 27% for HFmrEF. [3] Another prospective cohort study that included 4,495 AMI patients (2017–2019) reported a higher proportion (53.4%) of HFpEF among those patients who developed HF after AMI. In addition, it showed that patients with HFpEF had a higher risk of death compared to those without HF (median follow-up duration of one year: cumulative mortality 15.5% VS 4.2%). [4].
Presently, the available evidence suggests conventional HF therapeutic agents have no positive effect on improving the prognosis of AMI patients with HFpEF. [5], [6] Given the relative limitations of therapeutic strategies, it is of great importance to effectively identify those patients who are vulnerable to developing HFpEF after AMI and then manage them adequately to prevent or delay the occurrence of this complication. Previous studies have found some independent predictors for HFpEF occurring after AMI, such as high-sensitive C reactive protein, N-terminal pro-brain natriuretic peptide (NT-proBNP), multivessel disease (MVD), etc. [7] In clinical practice, nevertheless, if only a single predictor is relied upon or several predictors are subjectively combined based on empirical judgment, the accuracy of the prediction is often difficult to quantify.
Clinical prediction model, an objective and accurate prediction tool, is increasingly used in multiple investigative and clinical areas as a quantitative risk and benefit assessment tool. At present, only a limited number of clinical prediction models exist for assessing the risk of developing HFpEF in AMI patients. Hence, we aimed to establish such a feasible tool for early identification, hoping to provide assistance in improving AMI patients' prognosis.
2. Methods
2.1. Data source
In this retrospective study, we retrieved the relevant data from the Medical Information Mart for Intensive Care III (MIMIC-III, version 1.4) for the training cohort and from the electronic medical record of The First Affiliated Hospital of Jinan University between January and December 2022 for the external validation cohort.
MIMIC-III is a large, freely-accessible database containing related data associated with over 40,000 patients admitted to the Beth Israel Deaconess Medical Center (BIDMC) critical care units between 2001 and 2012. Available information includes demographics, laboratory examination results, imaging reports, diagnostic data, mortality, etc. The Institutional Review Boards of BIDMC and Massachusetts Institute of Technology approved this database project. A waiver of informed consent and the data sharing initiative were also authorized because the project did not impact clinical care, and all protected health information was deidentified. [8], [9] After completing a recognized course about the protection of human research participants and signing a data use agreement, we obtained permission to access the data in MIMIC-III for research purposes (Certification Number: 49272722). The Research Ethics Committee of The First Affiliated Hospital of Jinan University authorized a waiver of informed consent due to the retrospective observational character and approved this study (KY-2023–309).
2.2. Study population
Patients admitted to hospital as a result of AMI onset were eligible for inclusion if transthoracic echocardiography (TTE) showed their left ventricular ejection fraction (LVEF) ≥ 50%. We excluded individuals fulfilling the following criteria: (i) prior history of HF; (ii) considering that other etiologies caused HF; (iii) any one TTE result showing LVEF < 50% during hospitalization; or (iv) incomplete clinical data;
The diagnostic eligibility for patients from MIMIC-III in the training cohort was based on ICD-9 diagnosis codes (Supplementary I Table S1). In the external validation cohort, the diagnosis of AMI and HFpEF was in accordance with the Fourth Universal Definition of Myocardial Infarction [10] and the 2021 European Society of Cardiology HF Guidelines [11], respectively (Supplementary I Table S2). The included patients were grouped according to whether HFpEF occurred during hospitalization.
2.3. Data extraction
We collected the following information: 1) demographics: age, sex, body mass index (BMI); 2) comorbidities: hypertension, diabetes mellitus, hyperlipidemia, atrial fibrillation (AF), ventricular arrhythmia, cardiac shock, acute kidney injury (AKI). 3) first recorded vital signs after admission: systolic blood pressure (SBP), diastolic blood pressure (DBP), heart rate (HR); 4) first tested cardiac biomarkers after admission: cardiac troponin (cTn), creatine kinase-MB (CK-MB), NT-proBNP; 5) first tested hematology after admission: white blood cell count (WBC), red blood cells count (RBC), platelet (PLT), hemoglobin (HGB), hematocrit (HCT), neutrophil-to-lymphocyte ratio (NLR), red cell distribution width (RDW); 6) first tested blood biochemistry after admission: low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), total cholesterol (TC), triglyceride (TG), glycated hemoglobin (HbA1c), blood urea nitrogen (BUN), serum creatinine (Scr), lactate dehydrogenase (LDH), lactic acid, aspartate aminotransferase (AST), alanine transaminase (ALT), total bilirubin (TBIL), serum albumin, serum sodium, serum potassium, C-Reactive Protein (CRP); 7) others: MVD, myocardial infarct site.
2.4. Statistical analysis
Continuous variables conforming to normal distribution would be presented as mean ± standard deviation (SD) and compared using independent samples Student’s t-test, or otherwise would be presented as median [interquartile range (IQR)] and compared using the Mann–Whitney U test. For categorical variables, presentation as frequencies (percentages) and comparision using the chi-square test or Fisher’s exact test would be used. Variables with a missing proportion > 50% were eliminated from our study, and the other missing values would be imputed by multiple imputations using the MICE package of R software.
Univariate analysis was initially performed in the training cohort to screen for potential factors preliminarily. Variables with P < 0.2 in the univariate analysis were included in a multicollinearity analysis. Variance inflation factor (VIF) > 5 determines the existence of multicollinearity. [12] We then utilized LASSO regression to screen the variables further and incorporated them into the multivariate logistic regression analysis. Based on the results of LASSO-Logistic regression, a nomogram model for predicting the risk of HFpEF occurrence in AMI patients was constructed.
The model’s performance was assessed by discrimination [area under the receiver operator characteristic curve (AUC)], calibration capacity (calibration curve and Hosmer-Lemeshow test/ Brier score), and clinical usefulness [decision curve analysis (DCA)]. Internal validation of the model was conducted through 1000 bootstrap samples. Finally, The model was externally validated and updated (the “recalibration” method) using the validation cohort.
All data processes and statistical analysis were completed by R Studio (R 4.3.1). We considered a P-value of < 0.05 statistically significant.
3. Results
A total of 201 patients were finally included in the training cohort (34 HFpEF and 167 no-HF) and 65 in the validation cohort (44 HFpEF and 21 no-HF). The patients’ flow chart of our study is shown in Fig. 1. In the MIMIC-III database, study variables with a missing proportion > 50% included ProBNP, HbA1c, lactic acid, serum albumin, LDH, and CRP. Among the rest of the variables, the ratio of missing values ranges from 0% to 44.78% (Supplementary I Table S3), with an overall missing rate of 9.43%.
Fig. 1.
The patients’ flow chart of this study. AMI, acute myocardial infarction; LVEF, left ventricular ejection fraction; TTE, transthoracic echocardiography; HFpEF, heart failure with preserved ejection fraction; HF, heart failure.
3.1. Baseline characteristics of the included patients
The baseline characteristics of the investigated patients are depicted in Table 1. The results of the univariate analysis showed that except for age, hyperlipemia, AF, cardiac shock, AKI, cTn, WBC, NLR, BUN, and Scr, there were no statistically significant differences for the other variables (P > 0.05) between HFpEF group and no-HF group in the training cohort.
Table 1.
Baseline characteristics of the included patients in the training cohort and validation cohort.
| Variables | Training cohort | Validation cohort | |||
|---|---|---|---|---|---|
| HFpEF group (n = 34) | no-HF group (n = 167) | P-value | Total (n = 201) | Total (n = 65) | |
| Age (y) | 74 (64.5, 83) | 64 (55, 76) | 0.004 | 66 (56, 77) | 59 (50, 67) |
| Male (n%) | 17 (50) | 100 (60) | 0.382 | 117 (58) | 51 (78) |
| BMI (kg/m2) | 27.55 (23.16, 32.38) | 28 (24.97, 32.47) | 0.614 | 27.99 (24.91, 32.41) | 25.04 (23.44, 28.3) |
| SBP (mmHg) | 127.91 ± 23.99 | 128.87 ± 22.69 | 0.831 | 128.71 ± 22.86 | 125.45 ± 17.41 |
| DBP (mmHg) | 67.88 ± 18.72 | 67.87 ± 14.7 | 0.998 | 67.88 ± 15.4 | 74.91 ± 12.69 |
| HR (bpm) | 81.5 (72.75, 100.75) | 77 (68, 88) | 0.056 | 79 (68, 88) | 78 (71, 85) |
| Hypertension (n%) | 23 (68) | 115 (69) | 0.889 | 138 (69) | 36 (55) |
| Diabetes mellitus (n%) | 10 (29) | 47 (28) | 0.881 | 57 (28) | 25 (38) |
| Hyperlipemia (n%) | 11 (32) | 90 (54) | 0.036 | 101 (50) | 40 (62) |
| AF (n%) | 13 (38) | 33 (20) | 0.035 | 46 (23) | 2 (3) |
| Cardiac shock (n%) | 9 (26) | 11 (7) | 0.002 | 20 (10) | 2 (3) |
| Ventricular arrhythmias (n%) | 4 (12) | 16 (10) | 0.753 | 20 (10) | 2 (3) |
| AKI (n%) | 10 (29) | 12 (7) | < 0.001 | 22 (11) | 2 (3) |
| LDL(mg/dL) | 105.59 ± 34.25 | 99.98 ± 35.02 | 0.39 | 100.93 ± 34.87 | 121.63 ± 35.2 |
| HDL(mg/dL) | 45 (35.25, 55.75) | 41 (35, 55) | 0.435 | 42 (35, 55) | 42 (35, 46) |
| TC(mg/dL) | 175.62 ± 43.52 | 166.05 ± 39.69 | 0.242 | 167.67 ± 40.41 | 206.32 ± 55.11 |
| TG(mg/dL) | 112.5 (74.75, 194.5) | 114 (76, 179.5) | 0.946 | 113 (76, 180) | 126 (94, 175) |
| WBC(x109/L) | 12 (9.05, 14.85) | 10.2 (8, 12.4) | 0.041 | 10.5 (8.1, 13) | 9.42 (7.88, 11.24) |
| RBC(x1012/L) | 4.01 ± 0.82 | 4.25 ± 0.62 | 0.119 | 4.21 ± 0.66 | 4.77 ± 0.64 |
| RDW(%) | 13.65 (13.1, 14.38) | 13.4 (12.9, 14) | 0.252 | 13.4 (13, 14) | 13.23 (12.76, 13.63) |
| HGB (g/dL) | 12.05 (11, 14.38) | 13 (11.8, 14.2) | 0.145 | 12.77 ± 1.93 | 14.11 ± 1.77 |
| PLT(x109/L) | 233 (211.75, 278.25) | 238 (202.5, 296.5) | 0.735 | 236 (204, 288) | 228.6 (193, 267) |
| HCT(%) | 35.15 (32.95, 41.68) | 37.6 (34.3, 41.4) | 0.361 | 37.34 ± 5.32 | 42.50 ± 4.88 |
| NLR | 8.68 (6.72, 15.19) | 4.9 (2.44, 8.37) | < 0.001 | 5.8 (2.75, 9.36) | 3.72 (2.5, 5.93) |
| BUN (mg/dL) | 23 (16, 30.5) | 18 (14, 21) | < 0.001 | 18 (14, 22) | 15 (11, 17) |
| Scr (mg/dL) | 1.05 (0.83, 1.4) | 0.9 (0.8, 1.1) | 0.017 | 0.9 (0.8, 1.1) | 0.9 (0.8, 1.1) |
| ALT (IU/L) | 33 (20.75, 66.5) | 29 (20, 50) | 0.386 | 30 (20, 51) | 38 (27, 56) |
| AST (IU/L) | 47 (22, 125.25) | 44 (26, 84) | 0.996 | 44 (26, 89) | 51 (30, 115) |
| TBIL (mg/dL) | 0.4 (0.32, 0.6) | 0.5 (0.4, 0.8) | 0.071 | 0.5 (0.4, 0.8) | 0.74 (0.59, 0.96) |
| Serum sodium (mmol/L) | 139 (135.25, 140) | 139 (136, 141) | 0.275 | 139 (136, 141) | 140.6 (138.4, 142) |
| Serum potassium (mmol/L) | 4.2 (3.82, 4.85) | 4.1 (3.7, 4.4) | 0.101 | 4.1 (3.7, 4.4) | 3.85 (3.75, 4.08) |
| Multivessel disease (n%) | 26 (76) | 104 (62) | 0.167 | 130 (65) | 40 (62) |
| cTn(ng/mL) | 1.17 (0.1, 11.2) | 0.32 (0.04, 1.46) | 0.019 | 0.35 (0.05, 2.19) | 3.3 (0.51, 14) |
| CK-MB (ng/ml) | 22 (10.25, 109.5) | 17 (7, 58.5) | 0.171 | 18 (8, 59) | 34 (22, 66) |
| Infarct site (n%) | 0.292 | ||||
| inferior | 13 (38) | 78 (47) | 91 (45) | 9 (14) | |
| inferoposterior | 6 (18) | 18 (11) | 24 (12) | 6 (9) | |
| anterior | 3 (9) | 30 (18) | 33 (16) | 19 (29) | |
| Inferolateral | 2 (6) | 13 (8) | 15 (7) | 1 (2) | |
| lateral | 2 (6) | 4 (2) | 6 (3) | 0 (0) | |
| Anterolateral | 2 (6) | 4 (2) | 6 (3) | 0 (0) | |
| Unspecified | 6 (18) | 20 (12) | 26 (13) | 30 (46) | |
Abbreviations: HF, heart failure; HFpEF, heart failure with preserved ejection fraction; BMI, Body Mass Index; SBP, systolic blood pressure; DBP, diastolic blood pressure; HR, heart rate; AF, atrial fibrillation; AKI, acute kidney injury; LDL-C, low-density lipoprotein cholesterol; HDL, high density lipoprotein cholesterol; TC, total cholesterol; TG, triglyceride; WBC, white blood cell count; RBC, red blood cells count; RDW, red cell distribution width; HGB, hemoglobin; PLT, platelet; HCT, hematocrit; NLR, neutrophil-to-lymphocyte ratio; BUN, blood urea nitrogen; Scr, serum creatinine; ALT, alanine transaminase; AST, aspartate aminotransferase; TBIL, total bilirubin; cTn, cardiac troponin; CK-MB, creatine kinase-MB.
3.2. Development of prediction model in the training cohort
The variables with P < 0.2 in the univariate analysis included hyperlipidemia, AF, cardiac shock, AKI, MVD, age, HR, BUN, Scr, serum potassium, WBC, RBC, HGB, cTn, CK-MB, NLR, and TBIL. Considering the presence of multicollinearity among independent variables (Fig. 2A), LASSO regression with 5-fold cross-validation was applied to further variable selection. 9 variables with non-zero coefficients were retained by optimal lambda, including hyperlipidemia, cardiac shock, AKI, age, HR, BUN, CK-MB, NLR, and TBIL (Fig. 3).
Fig. 2.
(A) Test of multicollinearity after univariate Logistic regression analysis. Higher bars (>5) indicate collinearity issues. (B) Multivariate logistics regression analysis to determine the final predictors of the prediction model. AKI, acute kidney injury; AF, atrial fibrillation; HR, heart rate; BUN, blood urea nitrogen; Scr, serum creatinine; WBC, white blood cell count; NLR, neutrophil-to-lymphocyte ratio; TBIL, total bilirubin; RBC, red blood cells count; HGB, hemoglobin; cTn, cardiac troponin; CK-MB, creatine kinase-MB. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 3.
Variable selection by LASSO regression analysis. Lambda.1se was selected, corresponding to 9 variables with non-zero coefficients.
The results of final multivariate logistic regression manifested that age (OR 1.05, 95% CI 1.02–1.10, P = 0.0076), HR (OR 1.03, 95% CI 1.01–1.06, P = 0.0169), NLR (OR 1.16, 95% CI 1.06–1.29, P = 0.0022), and TBIL (OR 0.09, 95% CI 0.01–0.53, P = 0.0155) were the independent factors influencing the risk of new-onset HFpEF after AMI (Fig. 2B). Based on that, we constructed a prediction model with its equation as follows: P1 = explogit1/(1 + explogit1), where P1 is the predicted probability of developing HFpEF after AMI and logit1 = -7.74103 + 0.05158*age(year) + 0.02871*HR(bpm) + 0.19297*NLR − 2.33760*TBIL(mg/dL). The P1 value can be calculated automatically using Supplementary II. Moreover, a Nomogram plot for this prediction model is shown in Fig. 4.
Fig. 4.
Nomogram for estimation of HFpEF risk in AMI patients with acute myocardial infarction. NLR, neutrophil-to-lymphocyte ratio; TBIL, total bilirubin; HR, heart rate.
3.3. Evaluation of prediction model and internal verification
The prediction model yielded an AUC of 0.823 in the training cohort (Fig. 5A) and 0.804 in the Bootstrap internal verification. The calibration curve showed an excellent consistency of the model in the training cohort (Fig. 5B). The Hosmer–Lemeshow test (P = 0.183) also indicated that our model had a great calibration performance. In the DCA analysis (Fig. 5C), the blue solid line between the area of green dotted lines demonstrated that using the model to predict HFpEF risk could obtain net benefit while the threshold probability is between 7% and 98%.
Fig. 5.
The evaluation and internal verification of the prediction model. (A) ROC curve of the prediction model. (B) The calibration curve and Hosmer-Lemeshow of the prediction model. (C) The DCA curve of the prediction model. ROC, receiver operating characteristic. AUC, Area Under Roc Curve; DCA, decision curve analysis; HFpEF, heart failure with preserved ejection fraction.
3.4. External verification of prediction model and updating
The prediction model's performance was further verified externally by assessing its capability in the validation cohort. The results demonstrated that the AUC is 0.749 (95% CI 0.630–0.868), with a calibration slope of 0.833, calibration intercept of 3.746, and Brier score of 0.572 (Table 2). The original model was updated using the “recalibration” method by integrating more information from the validation cohort to make it more adaptable to the new patients.
Table 2.
The performance of differentiation and calibration in the validation cohort.
| Performance | Original model | Updated model |
|---|---|---|
| AUC (95% CI) | 0.749 (0.630–0.868) | 0.749 (0.630–0.868) |
| Calibration slope | 0.833 | 1 |
| Calibration intercept | 3.746 | 0 |
| Brier score | 0.572 | 0.185 |
Abbreviations: AUC, Area Under Roc Curve; CI, confidence interval;
According to the updated model, the predicted probability of developing HFpEF for AMI patients: P2 = explogit2/(1 + explogit2), where logit2 = -2.703911 + 0.04297561*age(y) + 0.0239207*HR(bpm) + 0.1607794*NLR − 1.94765*TBIL(mg/dL). Similarly, the P2 value can be calculated automatically using Supplementary II. As shown in Table 2, the calibration performance of the updated model in the validation cohort has improved significantly (slope 1, intercept 0, Brier score 0.185).
4. Discussion
The present study found that age, HR, NLR, and TBIL were the independent influencing factors for HFpEF occurrence in AMI patients. Our prediction model exhibited satisfying performance of discrimination, calibration, and clinical usefulness in evaluation and internal validation. The result of external verification showed that model differentiation has a moderate performance in the validation cohort, whereas calibration does not. Considering the discrepancy of regional and temporal between the training cohort and validation cohort, this result did not exceed our expectations and is consistent with previous research reports [13], [14], [15]. In this case, updating the original prediction model is generally recommended, which can be more suitable for the new cohort [13]. The recalibration consequence revealed that the updated model performed well and might be used to predict the probability of developing HFpEF for AMI patients in the validation cohort.
In addition, we used the new population data of the validation cohort to redevelop and revalidate a new prediction model. This methodological procedure is not suggested commonly, because the prior information captured in the original prediction model would be neglected when new patient data are used to develop a new model. [13] Nevertheless, this additional procedure's objective was to further verify the strength of association between 4 predictors and outcome rather than redevelop a new model. The redeveloped prediction model achieved good performance (AUC 0.827 95% CI 0.723–0.930, Hosmer-Lemeshow test P 0.993, calibration slope 1, calibration intercept 0, and Brier score 0.160), indirectly showing that the 4 independent predictors in the original model (i.e. age, HR, NLR and TBIL) were indeed strongly associated with outcome occurrence (i.e., developing HFpEF after AMI).
The increased prevalence of HFpEF with age has been found in previous studies [16], [17]. This is probably driven by aging-related cardiac mechanisms, such as increased inflammation and oxidative stress, mitochondrial dysfunction, and decreased autophagic capacity. Currently, there is evidence that some anti-aging interventions could have a therapeutic effect on HFpEF. For instance, anti-aging strategies controlled by diet were effective in improving diastolic dysfunction in elderly patients with HFpEF. [18] In an animal study, spermidine, a substance with extensive antiaging effects, prevented cardiac diastolic dysfunction in old mice by enhancing cardiomyocyte autophagy, inhibiting inflammation, and so on. [19] It is generally accepted that left ventricular diastolic dysfunction is the fundamental ingredient underlying the pathophysiology of HFpEF. [20] Another reason contributing to the association between age and HFpEF might be that older patients tend to have more comorbid cardiovascular risk factors, thus promoting HFpEF in a variety of facets. [16].
Patients with HFpEF usually display subtle abnormality in left ventricular systolic function. Despite a preserved LVEF, their cardiac output reserve and exercise tolerance might have decreased. For example, relatively mild systolic dysfunction at rest can deteriorate dramatically during exercise or physiological stress, resulting in the cardiac inability to meet the required cardiac output, thus producing symptoms of HF. [20] The reduced cardiac output mediated by damaged cardiomyocytes after AMI stimulates the initiation of the organism’s compensation to maintain effective circulating blood volume. Compared with patients without HF, patients with HFpEF may be more dependent on accelerated HR to increase ventricular stroke volume because of their cardiac incapability on enlarging left ventricular end-diastolic volume under the existence of diastolic dysfunction. [21] In consequence, HR augmentation could help early identification of patients who develop HFpEF after AMI.
Necrotic or apoptotic cardiomyocytes activate a severe inflammatory response through the danger-associated molecular patterns after AMI. [22] The inflammatory response then reduces nitric oxide (its mediated production of cyclic guanosine monophosphate could promote left ventricular diastole and inhibit myocardial hypertrophy) bioavailability in cardiomyocytes by inducing oxidative stress of coronary microvascular endothelial cells. [20] Therefore, inflammatory response represents a key pathophysiological mechanism for the development of HFpEF after AMI. As a novel inflammatory marker, NLR has been the focus of an increasing number of studies in recent years. A meta-analysis of 14 trials with 10,245 AMI patients found that the risk of major adverse cardiovascular events and death was significantly higher in the high NLR patients than in the low NLR patients. [23] The results of a prospective study showed that elevated NLR on admission was an independent risk factor for ventricular remodeling in patients with AMI. [24] So far, the association between NLR and HFpEF after AMI has not been reported, and this is the first study to illustrate it. We found high NLR is closely related to the occurrence risk of HFpEF in AMI patients.
Bilirubin was once considered as a simple metabolic end product of heme degradation but is now a crucial endogenous antioxidant molecule correlated with plenty of cardiovascular disease (CVD). [25] Many studies have concluded presently that appropriate elevation of TBIL levels can be protective against CVD. [26], [27], [28].
To confirm the effect of TBIL levels on HFpEF, Zheng et al. [25] conducted a retrospective study in which 327 patients with HFpEF were divided into four subgroups according to the severity of diastolic dysfunction measured by TTE and 200 healthy volunteers were recruited as a control group. The results showed that the TBIL levels were remarkably lower in the HFpEF group than in the healthy group despite adjusting for various covariates. Also, the more severe diastolic dysfunction patients had, the lower their TBIL levels were, which was shown in the subgroup analysis. Consistent with the above findings, our study likewise suggested a correlation between lower TBIL levels and a higher risk of developing HFpEF after AMI. The potential mechanism may be related to the fact that appropriately elevated TBIL levels produce the antioxidant effect, thereby attenuating responses such as inflammation-induced oxidative stress after AMI. [29].
This study has several limitations. First, this was a retrospective study with a relatively small sample size for inclusion, especially for the validation cohort (n = 65), which might inevitably introduce some bias and influence the statistical power. Therefore, the results should be interpreted with caution and future multicentre, larger prospective studies are needed to validate the findings of this study further. Second, due to limited data availability in the MIMIC-III database, certain variables of interest and clinically important data were not recorded or had a high proportion of missing values, and therefore could not be included in the model development. These variables include NT-proBNP, key echocardiography parameters (e.g., e’, E/e’, tricuspid regurgitation velocity, and left atrial volume index), culprit vessel, type of infarction, etc. Thus, we are looking forward to future research exploring whether there are any more favorable predictors. Third, multiple imputation was applied for TBIL and NLR due to relatively high missing proportions in the training cohort, which may have affected the reliability of the prediction model. Fourth, our prediction model based on the MIMIC-III database may have geographic, population, or temporal differences, leading to some limitations in the general applicability. However, our results showed the integrated model updated with new data upon the basis of the original model could be effectively applied to the new populations, suggesting that our prediction model holds particular value for practical clinical application. In other words, the updated model is expected to have improved general applicability and might potentially be suitable for cautious use in clinical practice. We anticipate further updates to the model by leveraging additional cohort data in the future, enabling it to be widely applicable to target populations across different regions.
5. Conclusions
In this study, we developed a preliminary prediction model incorporating four readily obtainable non-invasive clinical and biological indicators for assessing the risk of HFpEF occurrence in AMI patients. While the prediction model could potentially serve as a auxiliary tool for early identification of those high-risk individuals, our study has notable limitations that need to be addressed in future research.
Author Contributions.
Conception and design: Z.L. Acquisition and analysis of data: Z.L., Y.F.. Y.S., S.W., and S.F. Drafting of the manuscript: Z.L., Y.F., and Y.S. Critical revision of the manuscript for important intellectual content: N.B., D.C., and J.G. Statistical analysis: Z.L. Administrative, technical, or material support: Y.P., H.L., C.J., and J.W.
The data used and/or analyzed in the current study are available from the corresponding author, for reasonable uses, upon request.
Funding.
This study did not receive any funding.
CRediT authorship contribution statement
Zhijie Liu: Writing – original draft, Software, Methodology, Formal analysis, Data curation. Yiming Fan: Writing – original draft. Yiting Su: Writing – original draft. Shaorong Wu: Data curation. Sijun Feng: Data curation. Yun Pan: Resources. Hairui Li: Resources. Can Jiang: Resources. Jinlong Wang: Resources. Jun Guo: Writing – review & editing, Conceptualization. Dongdong Chen: Writing – review & editing, Conceptualization. Ning Bian: Writing – review & editing, Supervision, Conceptualization.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Footnotes
Supplementary data to this article can be found online at https://doi.org/10.1016/j.ijcha.2026.101917.
Contributor Information
Jun Guo, Email: dr.guojun@163.com.
Dongdong Chen, Email: 449244049@qq.com.
Ning Bian, Email: bian_ning@163.com.
Appendix A. Supplementary data
The following are the Supplementary data to this article:
Data availability
Data will be made available on request.
References
- 1.Rusu M., Hilse K., Schuh A., et al. Biomechanical assessment of remote and postinfarction scar remodeling following myocardial infarction. Sci. Rep. 2019;9(1):16744. doi: 10.1038/s41598-019-53351-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Desta L, Jernberg T, Löfman I, et al. Incidence, temporal trends, and prognostic impact of heart failure complicating acute myocardial infarction. The SWEDEHEART Registry (Swedish Web-System for Enhancement and Development of Evidence-Based Care in Heart Disease Evaluated According to Recommended Therapies): a study of 199,851 patients admitted with index acute myocardial infarctions, 1996 to 2008. JACC Heart Fail 2015; 3(3): 234-42. [DOI] [PubMed]
- 3.Kamon D., Sugawara Y., Soeda T., et al. Predominant subtype of heart failure after acute myocardial infarction is heart failure with non-reduced ejection fraction. ESC Heart Fail. 2020;8(1):317–325. doi: 10.1002/ehf2.13070. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Zhang X., Wang S., Sun L., Fang S., Yu B. Prognostic value of D-dimer in acute myocardial infarction complicated by heart failure with preserved ejection fraction. ESC Heart Fail. 2020;7(6):4118–4125. doi: 10.1002/ehf2.13027. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Song P.S., Kim M., Seong S.-W., et al. Heart failure with mid-range ejection fraction and the effect of β-blockers after acute myocardial infarction. Heart Vessels. 2021;36(12):1848–1855. doi: 10.1007/s00380-021-01876-1. [DOI] [PubMed] [Google Scholar]
- 6.Ferrari R. Effects of angiotensin-converting enzyme inhibition with perindopril on left ventricular remodeling and clinical outcome: results of the randomized Perindopril and Remodeling in elderly with Acute Myocardial Infarction (PREAMI) study. Arch. Intern. Med. 2006;166(6):659–666. doi: 10.1001/archinte.166.6.659. [DOI] [PubMed] [Google Scholar]
- 7.Xu M., Yan L., Xu J., Yang X., Jiang T. Predictors and prognosis for incident in-hospital heart failure in patients with preserved ejection fraction after first acute myocardial infarction: an observational study. Medicine (Baltimore) 2018;97(24) doi: 10.1097/MD.0000000000011093. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Johnson A., Pollard T., Mark R. MIMIC-III Clinical Database (version 1.4) [DB/OL] PhysioNet. 2016 doi: 10.13026/C2XW26. [DOI] [Google Scholar]
- 9.Johnson A.E.W., Pollard T.J., Shen L., et al. MIMIC-III, a freely accessible critical care database. Sci. Data. 2016;3 doi: 10.1038/sdata.2016.35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Thygesen K., Alpert J.S., Jaffe A.S., et al. Fourth Universal Definition of Myocardial Infarction (2018) Circulation. 2018;138(20):e618–e651. doi: 10.1161/CIR.0000000000000617. [DOI] [PubMed] [Google Scholar]
- 11.McDonagh T.A., Metra M., Adamo M., et al. 2021 ESC guidelines for the diagnosis and treatment of acute and chronic heart failure. Eur. Heart J. 2021;42(36):3599–3726. doi: 10.1093/eurheartj/ehab368. [DOI] [PubMed] [Google Scholar]
- 12.Kim J.H. Multicollinearity and misleading statistical results. Korean J. Anesthesiol. 2019;72(6):558–569. doi: 10.4097/kja.19087. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Janssen K.J.M., Moons K.G.M., Kalkman C.J., Grobbee D.E., Vergouwe Y. Updating methods improved the performance of a clinical prediction model in new patients. J. Clin. Epidemiol. 2008;61(1):76–86. doi: 10.1016/j.jclinepi.2007.04.018. [DOI] [PubMed] [Google Scholar]
- 14.Toll D.B., Janssen K.J.M., Vergouwe Y., Moons K.G.M. Validation, updating and impact of clinical prediction rules: a review. J. Clin. Epidemiol. 2008;61(11):1085–1094. doi: 10.1016/j.jclinepi.2008.04.008. [DOI] [PubMed] [Google Scholar]
- 15.Al-Farra H., de Mol B.A.J.M., Ravelli A.C.J., et al. Update and, internal and temporal-validation of the FRANCE-2 and ACC-TAVI early-mortality prediction models for Transcatheter Aortic Valve Implantation (TAVI) using data from the Netherlands heart registration (NHR. Int J Cardiol Heart Vasc. 2021;32 doi: 10.1016/j.ijcha.2021.100716. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Guo L., Guo X., Chang Y., et al. Prevalence and Risk Factors of Heart failure with Preserved Ejection Fraction: a Population-based Study in Northeast China. Int. J. Environ. Res. Public Health. 2016;13(8) doi: 10.3390/ijerph13080770. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Cui X., Zhou J., Jin X., et al. Prevalence and correlates of left ventricular diastolic dysfunction and heart failure with preserved ejection fraction in elderly community residents. Int. J. Cardiol. 2017;227:820–825. doi: 10.1016/j.ijcard.2016.10.041. [DOI] [PubMed] [Google Scholar]
- 18.Abdellatif M., Kroemer G. Heart failure with preserved ejection fraction: an age-related condition. J. Mol. Cell. Cardiol. 2022;167:83–84. doi: 10.1016/j.yjmcc.2022.03.008. [DOI] [PubMed] [Google Scholar]
- 19.Eisenberg T., Abdellatif M., Schroeder S., et al. Cardioprotection and lifespan extension by the natural polyamine spermidine. Nat. Med. 2016;22(12):1428–1438. doi: 10.1038/nm.4222. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Omote K., Verbrugge F.H., Borlaug B.A. Heart failure with Preserved Ejection Fraction: Mechanisms and Treatment strategies. Annu. Rev. Med. 2022;73:321–337. doi: 10.1146/annurev-med-042220-022745. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Bristow M.R., Altman N.L. Heart Rate in Preserved Ejection Fraction Heart failure. JACC Heart Fail. 2017;5(11):792–794. doi: 10.1016/j.jchf.2017.09.004. [DOI] [PubMed] [Google Scholar]
- 22.Oliveira J.B., Soares A.A.S.M., Sposito A.C. Inflammatory Response during Myocardial Infarction. Adv. Clin. Chem. 2018;84:39–79. doi: 10.1016/bs.acc.2017.12.002. [DOI] [PubMed] [Google Scholar]
- 23.Zhang S., Diao J., Qi C., et al. Predictive value of neutrophil to lymphocyte ratio in patients with acute ST segment elevation myocardial infarction after percutaneous coronary intervention: a meta-analysis. BMC Cardiovasc. Disord. 2018;18(1):75. doi: 10.1186/s12872-018-0812-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Börekçi A., Gür M., Türkoğlu C., et al. Neutrophil to Lymphocyte Ratio Predicts Left Ventricular Remodeling in patients with ST Elevation Myocardial Infarction after Primary Percutaneous Coronary intervention. Korean Circ J. 2016;46(1):15–22. doi: 10.4070/kcj.2016.46.1.15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Zheng H., Li Y., Xie N. Association of serum total bilirubin levels with diastolic dysfunction in heart failure with preserved ejection fraction. Biol. Res. 2014;47(1):7. doi: 10.1186/0717-6287-47-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Suh S., Cho Y.R., Park M.K., Kim D.K., Cho N.H., Lee M.-K. Relationship between serum bilirubin levels and cardiovascular disease. PLoS One. 2018;13(2) doi: 10.1371/journal.pone.0193041. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Wojciechowska C., Jacheć W., Romuk E., et al. Serum Sulfhydryl groups, Malondialdehyde, Uric Acid, and Bilirubin as Predictors of adverse Outcome in Heart failure patients due to Ischemic or Nonischemic Cardiomyopathy. Oxid. Med. Cell. Longev. 2021;2021 doi: 10.1155/2021/6693405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Lan Y., Liu H., Liu J., Zhao H., Wang H. Is serum total bilirubin a predictor of prognosis in arteriosclerotic cardiovascular disease? a meta-analysis. Medicine (Baltimore) 2019;98(42) doi: 10.1097/MD.0000000000017544. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Marconi V.C., Duncan M.S., So-Armah K., et al. Bilirubin is Inversely Associated with Cardiovascular Disease among HIV-Positive and HIV-Negative individuals in VACS (Veterans Aging Cohort Study) J. Am. Heart Assoc. 2018;7(10) doi: 10.1161/JAHA.117.007792. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data will be made available on request.