Prognostic value of the lactate-albumin ratio and development of a nomogram-based prediction model in patients with acute myocardial infarction–associated cardiogenic shock

Abstract

Cardiogenic shock (CS) is a serious complication of acute myocardial infarction (AMI), associated with in-hospital mortality rates of more than 40%. There is a lack of accuracy and clinical applicability in the existing risk scores, emphasizing the need for improved prognostic models using novel biomarkers. This study aimed to evaluate the prognostic role of the lactate-to-albumin ratio (LAR) in AMI-CS patients and to construct a nomogram-based model incorporating LAR with conventional clinical indicators. This prospective study enrolled 129 AMI-CS patients admitted to the cardiac care unit of Renmin Hospital (November 2022–December 2024). Patients were divided into survivor (n = 106) and in-hospital mortality (n = 23) groups. Clinical and laboratory data were compared using univariate analysis. Variables with prognostic potential were initially screened using Least Absolute Shrinkage and Selection Operator regression before entering multivariable logistic analysis. A nomogram was constructed using independent predictors and evaluated using receiver operating characteristic–area under the curve (AUC), calibration analysis, Hosmer–Lemeshow testing, and decision curve analysis. LAR, cold and clammy skin, aspartate aminotransferase, C-reactive protein, and left ventricular ejection fraction < 40% emerged as independent predictors of in-hospital mortality. The median (interquartile range) for LAR was significantly higher in the in-hospital mortality group versus the survival group (2.82 [0.93–4.2] vs 0.59 [0.42–1.54]), P = .002. The best cutoff value for LAR was 1.997, with 65% sensitivity and 82% specificity. LAR had an AUC of 0.767, outperforming blood lactate (0.749), and showed strong prognostic value (odds ratio: 1.495, 95% confidence interval: 1.174–1.904, P = .001). The nomogram-based model achieved an AUC of 0.928, outperforming CardShock (0.782), Acute Physiology and Chronic Health Evaluation II (0.727), and blood lactate (0.749). The model demonstrated excellent calibration (mean absolute error: 0.04), and goodness-of-fit was confirmed (Hosmer–Lemeshow P = .648). Decision curve analysis indicated superior clinical usefulness. LAR is a meaningful indicator for early mortality risk in AMI-CS. The proposed nomogram may support faster decision-making and bedside risk evaluation.

1. Introduction

Cardiovascular disease (CVD) are well known to be the leading cause of death, with recent estimates showing approximately 19.41 million deaths and over 612 million people affected globally.^[1] Acute myocardial infarction (AMI) is among the most severe coronary heart disease (CHD) presentations.^[2] Although only 3% to 13% of AMI cases result in cardiogenic shock (CS), the associated in-hospital mortality rate remains alarmingly high, reaching 40% to 50%.^[3]

In the last 20 years, despite advancements in pharmacological treatments and mechanical circulatory support, the rate of improvement in such patients has not been satisfactory.^[4] In-hospital mortality in AMI-CS patients is still high, even with optimized treatments. Therefore, identifying patients at high risk early is important for improving clinical outcomes. There has been increasing popularity of predictive tools since the beginning of the 1990s, but their full effect in clinical settings is not yet known.^[5] Several risk prediction scores have been developed for evaluating conditions and prognosis over the past decades, such as the Thrombolysis in Myocardial Infarction score, Sequential Organ Failure Assessment score, IABP-SHOCK II score, CardShock score, and Acute Physiology and Chronic Health Evaluation II (APACHE II) score.^[6]

However, the old risk scores more or less have shortcomings. Tools such as APACHE II and CardShock are commonly used but lack disease specificity, thus showing insufficient efficacy in AMI-CS patients. In addition, some scoring systems include indicators that lack early sensitivity, are insufficient to comprehensively reflect the body’s condition, or result in high detection costs, while other tools are complex to calculate and unfriendly to busy clinicians. Therefore, it is necessary to explore a more scientific scoring system that better fits the characteristics of cardiovascular critical illness and includes more accurate and easily obtainable new biomarkers to better guide clinical practice.

In recent years, interest has shifted toward novel biomarkers that offer improved risk predictive accuracy. Among them, the lactate-albumin ratio (LAR) has gained attention for its performance in critical illness. In sepsis, LAR has outperformed serum lactate alone in predicting in-hospital mortality.^[7] In cases of out-of-hospital cardiac arrest, LAR has also been linked to better neurologic outcomes and survival.^[8] Additionally, in heart failure (HF), LAR has shown promise in predicting both short- and long-term mortality.^[9]

New biomarkers such as LAR may be helpful for the prognosis of AMI-CS, but their use in clinical settings is still insufficient at present. This study examines how well LAR predicts in-hospital mortality in AMI-CS patients and builds a simple model using LAR and other clinical indicators to support early risk assessment and treatment decisions.

2. Materials and methods

2.1. Experimental design and participant inclusion

This prospective study was carried out from November 2022 to December 2024 in the Department of Cardiac Care Unit (CCU) of Renmin Hospital, Hubei University of Medicine, Shiyan, Hubei, China. The medical records of 129 AMI-CS patients admitted to the CCU were collected. In this study, data from 129 patients who met the diagnostic and inclusion criteria for AMI complicated by CS were analyzed, and complete baseline clinical and laboratory records were available. These were all consecutive patients admitted to the CCU during the study period. A separate screening record was not maintained; however, cases with missing important laboratory data or CS due to non-AMI causes were excluded based on the study criteria. Patients with incomplete or missing essential laboratory data were excluded during screening. A simple flow diagram shows how the patients were screened and included in the final analysis (see Fig. 1). After these exclusions, there were no missing values in the analyzed dataset. The study cases were divided into a survival group (n = 106) and an in-hospital mortality group (n = 23). All eligible subjects’ data were included in this study after informed consent was obtained.

Figure 1.

Flow diagram of patient selection and final study cohort. A formal screening log of excluded patients was not maintained, so detailed screening data are not available. AMI = acute myocardial infarction, CCU = cardiac care unit, CS = cardiogenic shock.

This study was conducted in accordance with the Declaration of Helsinki (1975, as revised in 2024) and was approved by the Institutional Review Board of Renmin Hospital, Hubei University of Medicine (Approval No. SYRMMY-2024-122). Written informed consent was obtained from the legally authorized representatives of all enrolled patients, in accordance with the procedure approved by the Institutional Review Board.

To minimize selection and information bias, uniform data collection procedures were applied throughout the study.

Society for Cardiovascular Angiography and Interventions shock classification was not performed because the required clinical information was not available.

The reporting of this study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines and the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis guidelines.^[10,11]

2.1.1. Inclusion criteria

• Age 18 years and above.

• Criteria meeting AMI: According to the fourth edition of the Global Definition of Myocardial Infarction criteria, MI is defined as acute myocardial injury (increased and/or decreased serum troponin and at least 1 occasion above the 99th percentile of the upper limit of the normal reference value) with evidence of acute myocardial ischemia, including symptoms of acute myocardial ischemia, new ischemic electrocardiogram changes, pathological Q-wave formation, imaging studies showing new myocardial necrosis or abnormal local wall movement consistent with ischemic etiology, and intracoronary thrombosis confirmed by coronary angiography or intracavitary imaging.^[12]

• Criteria meeting CS: hypotension (systolic blood pressure [SBP] < 90 mm Hg or the need for vasopressors to maintain SBP ≥ 90 mm Hg) accompanied by signs of organ hypoperfusion, such as altered mental status, liver dysfunction, renal dysfunction (urine output < 30 mL/h), and elevated serum lactate levels (>2 mmol/L) in the presence of cardiac dysfunction.^[13]

2.1.2. Exclusion criteria

• Age < 18 years.

• CS due to non-AMI causes, such as valve diseases, cardiomyopathy, and myocarditis.

• Patients who died during the occurrence of CS, resulting in critical data loss.

• Patients with mental, cognitive, and behavioral disorders.

• Patients with primary blood system diseases, infections, and immune system diseases.

• Patients who have recently taken glucocorticoids or immunosuppressants.

• Patients with recent trauma.

• Patients with missing laboratory results or values.

2.2. Data collection and computation

2.2.1. Clinical data

General data (age, gender, and elderly), past medical history (hypertension, type 2 diabetes mellitus [T2DM], CHD, HF, atrial fibrillation, cerebral hemorrhage, and smoking), vital signs at admission to the CCU (heart rate [HR], blood oxygen saturation [SpO₂], diastolic blood pressure, mean arterial pressure, and SBP), laboratory investigations (left ventricular ejection fraction [LVEF], white blood cell count [WBC], hemoglobin [Hb], platelets, absolute neutrophil count, absolute lymphocyte count, red blood cell distribution width, neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, serum sodium [Na⁺], serum potassium [K⁺], serum calcium [Ca²⁺], pH, serum bicarbonate [HCO₃⁻], blood lactate, C-reactive protein [CRP], serum creatinine, estimated glomerular filtration rate [eGFR], blood urea nitrogen [BUN], cystatin C, uric acid, alanine aminotransferase [ALT], aspartate aminotransferase [AST], albumin, triglyceride, fasting blood sugar, high-density lipoprotein [HDL], low-density lipoprotein, cardiac troponin I, and N-terminal pro-B-type natriuretic peptide [NT-proBNP]), data related to cardiac problems (ST-elevation myocardial infarction [STEMI], non-STEMI, emergency percutaneous coronary intervention [PCI], and new onset of atrial fibrillation), medications and treatment procedures (intra-aortic balloon pump [IABP], extracorporeal membrane oxygenation, use of 2 or more vasopressors, continuous renal replacement therapy, mechanical ventilation, defibrillation, and albumin infused or not), and other related data (confusion, cold and clammy skin, urine output, whether it was secondary hospitalization, and whether the patient survived or died). Hospital medical records were accessed for data collection. All significant laboratory tests, such as blood lactate, albumin, AST, and CRP, were collected at the time of admission to the CCU, as close as possible to when CS was diagnosed, and before treatment if possible.

2.2.2. Calculation of relevant parameters

1. Triglyceride-glucose index (TyG index) = Ln (fasting triglyceride [mg/dL] × fasting glucose [mg/dL]/2); (triglycerides: 1 mmol/L = 88.6 mg/dL; fasting blood glucose: 1 mmol/L = 18 mg/dL)

2. LAR = lactate (mmol/L)/albumin (g/L)

3. Serum creatinine-to-cystatin C ratio = serum creatinine (mg/dL)/cystatin C (mg/L) × 10

4. AST-to-ALT ratio = AST (U/L)/ALT (U/L)

5. AIP = log₁₀ (triglycerides [mg/dL]/HDL cholesterol [mg/dL]); (triglycerides: 1 mmol/L = 88.6 mg/dL; HDL cholesterol: 1 mmol/L = 38.67 mg/dL)

6. Shock index (SI) = HR (bpm)/SBP (mm Hg)

7. Reverse SI = SBP (mm Hg)/HR (bpm)

8. Neutrophil-percentage-to-albumin ratio = (neutrophil percentage (%) × 100)/albumin (g/dL)

9. APACHE II = acute physiology score + age points + chronic health points (minimum score = 0; maximum score = 71)

10. CardShock = age (0–3 points) + previous myocardial infarction (MI) or coronary artery bypass grafting (CABG; 0–1 points) + acute coronary syndrome (ACS) etiology (0–1 points) + LVEF < 40% (0–1 points) + serum lactate (mmol/L) (0–2 points) + eGFR (mL/min/1.73 m²) (0–2 points) + confusion (0–1) (risk categories: low risk = 0–3 points; intermediate risk = 4–5 points; high risk = 6–9 points)

2.3. Statistical analysis

The baseline characteristics of the study cases were described using descriptive analysis, with categorical variables presented as frequencies (n) and percentages (%). The normality of the variables was assessed using the Shapiro–Wilk test. Normally distributed continuous variables were presented as mean ± standard deviation (x̄ ± s), and the median and interquartile range were used for unevenly distributed variables. Categorical variables were analyzed using the chi-square test (χ²) or Fisher’s exact test. An independent-samples t test was used for normally distributed continuous variables, and the Mann–Whitney U test was used for unevenly distributed continuous variables. Clinically significant variables were selected as the main indicators for multivariable logistic regression. Least Absolute Shrinkage and Selection Operator (LASSO) penalized regression was used to choose the variables because the number of deaths was small. The penalty (lambda) was set by 10-fold cross-validation, and only variables with nonzero coefficients were retained for the final model. Finally, 5 main variables were selected to build the prediction model. The nomogram was also constructed from the main selected indicators. The predictive accuracy and performance of the nomogram were evaluated using the nomogram calibration curve. Internal validation of the final model was performed using bootstrap resampling with 1000 repetitions to calculate the optimism-corrected area under the receiver operating characteristic (ROC) curve (AUC) and the calibration slope. Univariate regression was performed for LAR to assess its predictive value. The receiver operating characteristic–area under the curve (ROC-AUC) values of the prediction model, LAR, blood lactate, APACHE II, and CardShock scores were compared with each other. A Hosmer–Lemeshow goodness-of-fit test was performed for the prediction model to evaluate model fit. P-values < .05 were considered significant. Decision curve analysis (DCA) compared the nomogram with APACHE II and CardShock across different threshold probabilities to assess net clinical benefit. Net benefit curves were used to determine whether the model performed better than standard scores and default strategies.

For the analysis, SPSS 27.0.0.0 (IBM Corp.) was used for univariate analysis, the Hosmer–Lemeshow goodness-of-fit test, and ROC-AUC analysis. In R 4.2.2 (R Foundation for Statistical Computing), the “glmnet” package was used for LASSO regression, and the “rms” package was used for nomogram construction. Bootstrap internal validation with 1000 resamples and DCA were performed using the “rms” package to obtain the optimism-corrected AUC and calibration slope.

No sensitivity analysis was performed in this study.

3. Results

3.1. Clinical characteristics of the patients

Comparison of the general data and all the baseline characteristics of the cases in the study was performed after dividing them into a survival group (n = 106) and an in-hospital mortality group (n = 23). The in-hospital mortality group had a higher proportion of LVEF < 40%, IABP, extracorporeal membrane oxygenation, use of 2 or more vasopressors, continuous renal replacement therapy, mechanical ventilation, defibrillation, confusion, and cold and clammy skin compared with the survival group (P < .05). Similarly, WBC, absolute neutrophil count, blood lactate, CRP, serum creatinine, uric acid, AST, ALT, LAR, SI, and neutrophil-percentage-to-albumin ratio were higher in the in-hospital mortality group than in the survival group (P < .05). However, secondary hospitalization or not, pH, HCO₃⁻, eGFR, albumin, and reverse SI were lower in the in-hospital mortality group compared with the survival group (P < .05; see Table 1).

Table 1.

Baseline clinical characteristics.

Variables	Survival (n = 106)	In-hospital mortality (n = 23)	P-value
	x̄ ± s, median (IQR), n (%)	x̄ ± s, median (IQR), n (%)
General data
Elderly (≥65 yr)	70 (66.04)	18 (78.26)	.254
Gender (male)	65 (61.32)	17 (73.91)	.255
Age (yr)	66.55 ± 13.03	67.65 ± 13.85	.716
Past medical history
History of HTN	54 (50.94)	10 (43.48)	.516
History of T2DM	27 (25.47)	2 (8.70)	.081
History of CHD	35 (33.02)	7 (30.43)	.811
History of HF	19 (17.92)	5 (21.74)	.768
Previous history of AF	4 (3.77)	0 (0.00)	1.000
History of stroke	14 (13.21)	3 (13.04)	1.000
Smoking history	39 (36.79)	10 (43.48)	.549
Vital signs during admission in CCU
HR (beats/min)	88.5 (73.75–104.25)	100 (79–119)	.15
SpO2 (%)	97.5 (95–98)	97 (94–98)	.659
SBP (mm Hg)	96 (86–115)	85 (79–108)	.090
DBP (mm Hg)	60 (53–70)	61 (47–70)	.362
MAP (mm Hg)	71.67 (64.75–85)	68 (58.33–81.67)	.187
Laboratory investigations
LVEF < 40%	25 (23.58)	14 (60.87)	<.001*
WBC (109/L)	9.71 (8.06–14)	15.12 (10.26–19.43)	.007*
Hb (g/L)	123.92 ± 22.69	118.35 ± 22.68	.287
Platelets (109/L)	203.34 ± 66.42	222.74 ± 79.19	.223
Absolute neutrophil count (109/L)	7.85 (5.88–10.73)	13.32 (7.01–16.21)	.009*
Absolute lymphocyte count (109/L)	1.09 (0.83–1.72)	1.19 (0.83–2.01)	.909
RDW (fL)	43.75 (41.58–47.05)	45.40 (41.60–83.60)	.256
NLR	6.76 (4.24–10.48)	8.21 (4.44–15.78)	.196
PLR	147.80 (112.05–257.97)	180 (94.49–234.11)	.858
Na+ (mmol/L)	138 (135.68–141.00)	139.7 (136.1–140.2)	.456
K+ (mmol/L)	3.78 (3.5–4.1)	3.83 (3.37–4.39)	.958
Ca2+ (mmol/L)	1.04 (0.94–1.11)	1.05 (0.94–1.09)	.567
pH	7.39 (7.34–7.45)	7.34 (7.19–7.44)	.006*
HCO3− (mmol/L)	21 (18–24)	17.8 (16–20.7)	.002*
Blood lactate (mmol/L)	2.04 (1.58–3.73)	6.17 (2.53–12)	<.001*
CRP (mg/L)	5.85 (1.83–34.32)	37.52 (5.48–87.77)	.006*
Serum creatinine (μmol/L)	89.6 (72.4–120.88)	111.8 (89–138.6)	.005*
eGFR (mL/min/1.73 m2)	76.1 (51.4–90.81)	60.08 (43.22–70.14)	.011*
BUN (mmol/L)	6.83 (5.69–9.67)	8.21 (5.49–13.85)	.162
Cystatin C (mg/L)	1.21 (0.93–1.69)	1.50 (1.11–2.22)	.064
UA (μmol/L)	322.2 (275.35–421.68)	419 (309–581.3)	.013*
AST (U/L)	41 (22–156.25)	233 (45–640.5)	.001*
ALT (U/L)	28.5 (16–54)	96 (21.1–212)	.003*
Albumin (g/L)	38.75 ± 4.99	33.58 ± 5.06	<.001*
TG (mmol/L)	1.25 (0.83–1.85)	1.24 (0.93–2.30)	.584
FBS (mmol/L)	8.50 (6.55–11.62)	8.11 (5.82–12.94)	.664
HDL (mmol/L)	0.97 (0.79–1.13)	0.81 (0.60–1.01)	.023*
LDL (mmol/L)	2.47 (1.90–3.09)	2.33 (1.61–3.32)	.256
cTnI (ng/mL)	2.08 (0.67–4.64)	4.30 (1.26–10)	.061
NT-proBNP (pg/mL)	1944.50 (382.03–8439.50)	4187.00 (1295–18,126)	.055
Data related to cardiac problems
STEMI	70 (66.04)	17 (73.91)	.465
NSTEMI	36 (33.96)	6 (26.09)	.465
Emergency PCI	49 (46.23)	6 (26.09)	.077
New onset of AF	13 (12.26)	3 (13.04)	1.000
Medications and treatment procedures
IABP	10 (9.43)	6 (26.09)	.039*
ECMO	6 (5.66)	5 (21.74)	.026*
Use of 2 or more vasopressors	35 (33.02)	13 (56.52)	.035*
CRRT	6 (5.66)	5 (21.74)	.026*
Mechanical ventilation	28 (26.42)	17 (73.91)	<.001*
Defibrillation	9 (8.49)	8 (34.78)	.003*
Albumin infused or not	2 (1.89)	1 (4.35)	.448
Other related data
Confusion	29 (27.36)	18 (78.26)	<.001*
Cold and clammy skin	33 (31.13)	19 (82.61)	<.001*
Urine output	37 (34.91)	13 (56.52)	.054
Secondary hospitalization or not	44 (41.51)	1 (4.35)	.001*
New biomarkers or indicators
TyG index	4.82 (4.47–5.09)	4.75 (4.41–5.12)	.834
LAR	0.59 (0.42–1.54)	2.82 (0.93–4.20)	.002*
CCR	8.02 (6.80–9.87)	8.91 (7.02–11.20)	.25
AST-ALT ratio	1.66 (1.22–3.05)	2.20 (1.59–3.37)	.169
AIP	0.10 (−0.12 to 0.27)	0.23 (−0.05 to 0.54)	.086
SI	0.87 (0.70–1.08)	1.04 (0.80–1.20)	.042*
rSI	1.15 (0.93–1.42)	0.96 (0.83–1.25)	.042*
NPAR	1.93 (1.55–2.99)	3.53 (2.38–4.95)	<.001*

Data presented as n (%), mean ± SD (x̄ ± s), and median (IQR).

AF = atrial fibrillation, AIP = atherogenic index of plasma, ALT = alanine aminotransferase, AST = aspartate aminotransferase, AST-ALT ratio = aspartate aminotransferase-alanine aminotransferase ratio, BUN = blood urea nitrogen, Ca²⁺ = serum calcium, CCR = serum creatinine-to-cystatin C ratio, CCU = cardiac care unit, CHD = coronary heart disease, CRP = C-reactive protein, CRRT = continuous renal replacement therapy, cTnI = cardiac troponin I, DBP = diastolic blood pressure, ECMO = extracorporeal membrane oxygenation, eGFR = estimated glomerular filtration rate, FBS = fasting blood sugar, Hb = hemoglobin, HCO₃⁻ = serum bicarbonate, HDL = high-density lipoprotein, HF = heart failure, HR = heart rate, HTN = hypertension, IABP = intra-aortic balloon pump, IQR = interquartile range, K⁺ = serum potassium, LAR = lactate-albumin ratio, LDL = low-density lipoprotein, LVEF = left ventricular ejection fraction, MAP = mean arterial pressure, Na⁺ = serum sodium, NLR = neutrophil-to-lymphocyte ratio, NPAR = neutrophil-percentage-to-albumin ratio, NSTEMI = non-ST-elevation myocardial infarction, NT-proBNP = N-terminal pro-B-type natriuretic peptide, PCI = percutaneous coronary intervention, PLR = platelet-to-lymphocyte ratio, RDW = red blood cell distribution width, rSI = reverse shock index, SBP = systolic blood pressure, SI = shock index, SPO₂ = blood oxygen saturation, STEMI = ST-elevation myocardial infarction, T2DM = type 2 diabetes mellitus, TG = triglyceride, TyG index = triglyceride-glucose index, UA = uric acid.

^*P < .05 between the survival group and the in-hospital mortality group.

3.2. Association of LAR with in-hospital mortality

The median (interquartile range) of LAR in the in-hospital mortality group was higher than that of the survival group (2.82 [0.93–4.2] vs 0.59 [0.42–1.54]). The odds ratio, 95% confidence interval [CI], and P-value were 1.495, 1.174 to 1.904, and .001, respectively, in univariate logistic regression, showing prognostic value for in-hospital mortality in AMI-CS patients (see Table 2).

Table 2.

Comparison of median LAR values between survivor and in-hospital mortality groups, and univariate logistic regression demonstrating its association with mortality in AMI-CS patients.

Variable	Survival (n = 106)	In-hospital mortality (n = 23)	Univariate regression
	Median (IQR)	Median (IQR)	95% CI	OR	P-value
LAR	0.59 (0.42–1.54)	2.82 (0.93–4.2)	1.174–1.904	1.495	.001

AMI = acute myocardial infarction, CI = confidence interval, CS = cardiogenic shock, IQR = interquartile range, LAR = lactate-albumin ratio, OR = odds ratio.

3.3. Selection of indicators and development of prediction model

From the LASSO analysis, 18 significant indicators showed nonzero coefficients at an optimal lambda (λ) of 0.032 (see Figs. 2 and 3). Among the 18 indicators, LVEF < 40%, cold and clammy skin, CRP, AST, and LAR, with nonzero coefficients of 0.3541, 0.7730, 0.0073, 0.0015, and 0.0507, respectively, were selected as the main indicators to create the prediction model.

Figure 2.

Ten-fold cross-validation curve showing optimal λ based on minimum deviance and 1-SE criteria. SE = standard error.

Figure 3.

Coefficient paths of all predictors across log(λ), with variables retained beyond shrinkage for model construction.

3.4. Multivariable logistic regression analysis results

In the multivariable logistic regression analysis of AMI-CS patients, LAR, cold and clammy skin, AST, CRP, and LVEF < 40% were selected as main variables. These variables were chosen by LASSO regression. The result showed that LAR, cold and clammy skin, AST, CRP, and LVEF < 40% were significant independent risk factors for in-hospital mortality with P < .05 (see Table 3).

Table 3.

Multivariable logistic regression analysis showing risk factors influencing in-hospital mortality in AMI-CS patients.

Variables	B	SE	Wald	P-value	OR	95% CI
LAR	0.493	0.162	9.301	.002	1.637	1.193–2.248
Cold and clammy skin	2.309	0.739	9.755	.002	10.067	2.364–42.881
AST	0.004	0.002	7.241	.007	1.004	1.001–1.007
CRP	0.014	0.005	9.134	.003	1.014	1.005–1.023
LVEF < 40%	1.355	0.652	4.318	.038	3.876	1.080–13.908
Constant	−6.031	1.133	28.348	<.001	0.002

AMI = acute myocardial infarction, AST = aspartate aminotransferase, CI = confidence interval, CRP = C-reactive protein, CS = cardiogenic shock, LAR = lactate-albumin ratio, LVEF = left ventricular ejection fraction, OR = odds ratio, SE = standard error.

3.5. Equation to calculate prediction of in-hospital mortality

$$\begin{array}{cc}{\displaystyle }\mathrm{L}\mathrm{o}\mathrm{g}\mathrm{i}\mathrm{t}\left(P\right)=& -6.031+0.493\left(\mathrm{L}\mathrm{A}\mathrm{R}\right)\\ & +2.309\left(\mathrm{c}\mathrm{o}\mathrm{l}\mathrm{d}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{c}\mathrm{l}\mathrm{a}\mathrm{m}\mathrm{m}\mathrm{y}\mathrm{s}\mathrm{k}\mathrm{i}\mathrm{n}\right)\\ & +0.004\left(\mathrm{A}\mathrm{S}\mathrm{T}\right)+0.014\left(\mathrm{C}\mathrm{R}\mathrm{P}\right)\\ & +1.355(\mathrm{L}\mathrm{V}\mathrm{E}\mathrm{F}<40\%)\end{array}$$

3.6. ROC-AUC analysis results of prediction model, LAR, blood lactate, CardShock, and APACHE II

The ROC-AUC analysis of the prediction model, LAR, blood lactate, CardShock, and APACHE II showed that the prediction model had the highest discriminative ability with an AUC of 0.928 (95% CI: 0.882–0.975), sensitivity of 82.6%, and specificity of 91.5%. The LAR cutoff from our data showed good predictive value with an AUC of 0.767 (95% CI: 0.656–0.878), sensitivity of 65.2%, and specificity of 82.1%, which was almost the same as that of CardShock (AUC 0.782). Blood lactate and APACHE II showed lower AUCs of 0.749 and 0.727, respectively (see Table 4 and Fig. 4). LAR was also better than blood lactate alone, which means the ratio may give more useful information than blood lactate by itself.

Table 4.

ROC-AUC analysis results for the prediction model, CardShock, LAR, blood lactate, and APACHE II.

Variables	AUC	P-value	95% CI	Best cutoff point	Sensitivity	Specificity	Youden index
Prediction model	0.928	<.001	0.882–0.975	0.337	0.826	0.915	0.741
CardShock	0.782	<.001	0.681–0.884	4.5	0.652	0.811	0.463
LAR	0.767	<.001	0.656–0.878	1.977	0.652	0.821	0.473
Blood lactate	0.749	<.001	0.618–0.879	8.54	0.478	0.953	0.431
APACHE II	0.727	.001	0.598–0.857	27.5	0.522	0.962	0.484

APACHE II = Acute Physiology and Chronic Health Evaluation II, AUC = area under the curve, CI = confidence interval, LAR = lactate-albumin ratio, ROC-AUC = receiver operating characteristic–area under the curve.

Figure 4.

ROC curves comparing the nomogram with LAR, blood lactate, CardShock, and APACHE II for predicting in-hospital mortality. APACHE II = Acute Physiology and Chronic Health Evaluation II, AUC = area under the curve, LAR = lactate-albumin ratio, ROC = receiver operating characteristic.

3.7. Nomogram development from the selected indicators and calibration curves assessment

LAR, cold and clammy skin, AST, CRP, and LVEF < 40% were incorporated to construct the final nomogram model (see Fig. 5).

Figure 5.

Nomogram for estimating in-hospital mortality risk in AMI-CS patients. AMI = acute myocardial infarction, AST = aspartate aminotransferase, CRP = C-reactive protein, CS = cardiogenic shock, LAR = lactate-albumin ratio, LVEF = left ventricular ejection fraction.

The nomogram calibration curve was used to show the effectiveness of the nomogram, which showed good consistency, where the bias-corrected line was close to the ideal line, and the mean absolute error was 0.04, showing that the model’s prediction was reliable (see Fig. 6).

Figure 6.

Calibration curve for predicting in-hospital mortality showing good alignment between predicted and observed risk.

The bootstrap internal validation (1000 resamples) yielded an optimism-corrected AUC of 0.904 and a calibration slope of 0.764, indicating good discrimination and calibration of the model. The calibration curve after 1000 bootstrap resamples showed that the bias-corrected line was close to the ideal diagonal, demonstrating good agreement between the predicted and observed probabilities (Fig. 7).

Figure 7.

Bootstrap-adjusted calibration curve (1000 resamples) demonstrating consistent model performance.

The Hosmer–Lemeshow goodness-of-fit test showed a P-value of .648, showing that there was no significant difference between observed and predicted probabilities.

3.8. DCA of nomogram compared with that of APACHE II and CardShock

DCA suggested that the nomogram offered greater net clinical benefit across a wide range of threshold probabilities compared with CardShock and APACHE II (Fig. 8). The nomogram curve consistently lay above the other models and the default strategies, showing that it was useful for guiding mortality-related clinical decisions.

Figure 8.

Decision curve analysis comparing the net clinical benefit of the nomogram with CardShock and APACHE II across a range of threshold probabilities. APACHE II = Acute Physiology and Chronic Health Evaluation II.

4. Discussion

In this study of 129 AMI-CS patients, LAR level was higher in the in-hospital mortality group compared with the survival group, suggesting its prognostic value in AMI-CS. The ROC-AUC (0.767), sensitivity (65.2%), and specificity (82.1%) also indicate predictive usefulness. In our data, LAR performed better than lactate alone, implying that it may be a more reliable marker.

As widely used and readily accessible clinical biomarkers, arterial blood lactate reflects tissue hypoperfusion and anaerobic metabolism, whereas albumin levels indicate the body’s inflammatory response and nutritional status. Numerous studies have demonstrated that hyperlactatemia and hypoalbuminemia are significantly associated with adverse outcomes in patients with STEMI.^[14,15] The composite indicator LAR, which integrates multiple pathophysiological dimensions, including metabolic stress and systemic inflammatory-nutritional status, may offer a more sensitive prognostic assessment for critically ill patients. Previous studies have shown that increased LAR is associated with poor prognosis in patients with CVD. Our findings are highly consistent with this conclusion. Also, instead of only using either lactate or albumin or a combination of both, LAR was superior for the prognosis of cardiac arrest during hospitalization along with good neurological results.^[16] In cases of septic MI, increased levels of LAR are also seen as an indication of the patient being at risk of in-hospital mortality.^[17] A recent study showed that increased LAR indicates increased chances of mortality in AMI patients, where metabolic stress and systemic inflammation become the main causes.^[18] There is also an increased risk of mortality while predicting 28-day mortality in cases of CHD with increased LAR.^[19]

Many studies regarding CVD, coronary artery disease, major adverse cardiovascular events, AMI, HF, and other related diseases have already been conducted using LAR alone, but studies of combining LAR as a new biomarker with traditional indicators have not been conducted in AMI-CS patients. Zhu et al^[20] studied LAR combined with HCO₃⁻ in AMI cases and showed that higher levels were associated with higher risks of death.

In CVD patients, the already present lab values, previous medical history, and risk factors help in the diagnosis and treatment of the disease. However, with these indicators, risk prediction tools are also made that can easily predict the risk of death, survival chances, and other outcomes.^[21] The APACHE II risk score, which was first proposed in 1984, is a modified form of APACHE, which includes common parameters such as HR, respiratory rate, mean blood pressure, body temperature, Glasgow Coma Scale, WBC, Na⁺, K⁺, serum creatinine, hematocrit, pH, and partial pressure of oxygen.^[22] Ludwig and Hulting^[23] in 1995 proposed that APACHE II can effectively predict the risk of mortality in intensive care unit patients who had AMI previously. A recent study showed that APACHE II had good predictive ability for mortality risk in ACS patients compared with inflammatory markers, with a sensitivity of 70.4% and a specificity of 92.9%.^[24] Related to the short-term death rates in patients with ACS as well as those not having ACS complicated with CS, the CardShock risk score was developed, which included common parameters mostly related to CS, such as age, lactate, eGFR, LVEF < 40%, ACS causes, and whether there was a history of CABG or MI.^[25]

In this study, a prediction model was developed by combining LAR with cold and clammy skin (a sign of peripheral hypoperfusion), AST, CRP, and LVEF < 40%, and was compared with APACHE II and CardShock. The model performed better than both scoring systems, with an AUC of 0.928 (95% CI: 0.882–0.975, P < .001). CardShock had an AUC of 0.782, and APACHE II had an AUC of 0.727. Internal bootstrap validation demonstrated good stability (optimism-corrected AUC = 0.904; calibration slope = 0.756), supporting the model’s reliability.

In CVD, the establishment of predictive tools, risk scores, and nomograms with common indicators to predict the risk of death in AMI patients is frequently done. Tang et al^[26] established a nomogram using variables such as the age of the patient, BUN, respiratory rate, Hb, presence of pneumonia, CS, dialysis, and use of mechanical ventilation. Being superior to Sequential Organ Failure Assessment and simplified acute physiology score II, it is able to predict survival of AMI patients. Another risk score called Santiago Shock Score (S3) was made with parameters such as mitral valve regurgitation (level 3–4), age, lactate, ACS causes, Hb, and NT-proBNP. It was able to predict the risk of death in CS cases with an AUC of 0.85, more than that of CardShock (AUC = 0.74).^[27] Yamga et al^[28] designed the BOS and MA2 risk scores with variables such as maximum BUN (25 mg/dL), SpO₂ (<88%), SBP (<80 mm Hg), use of mechanical ventilation, age (limit = 60 years), and anion gap of 14 mmol/L for AMI-CS patients. It also showed efficient performance results with an AUC of 0.83 in the training and 0.76 in the validation set.

The study of new biomarkers is needed because such biomarkers can detect when CS is starting, progression of CS, severity, detection of end-organ hypoperfusion, MODS, and with such information, early management of CS is possible.^[29] Liu et al^[30] studied the TyG index in cases of AMI developing CS and showed that patients who developed CS had a higher TyG index, indicating an association of the TyG index with AMI-CS risk prediction. Schupp et al^[31] studied CS patients and 30-day mortality with the AST-ALT ratio along with bilirubin and showed the increased chances of death of the patients in the intensive care unit if the AST-ALT ratio is high, where the AST-ALT ratio is observed on day 4. Colarusso et al^[32] proposed that SI with its variants, such as adjusted SI, age-modified SI, and SI-C, showed good prognostic value in mortality prediction in CS patients.

Recently, Ceglarek et al^[33] in their study developed a risk score called the CLIP score using commonly used biomarkers in CVD, which are lactate and NT-proBNP, and also added cystatin C and interleukin-6, which are new in CVD risk score studies, for mortality prediction in AMI-CS patients. The CLIP score outperformed simplified acute physiology score II, IABP-SHOCK II, and SHOCK Trial risk scores with AUC values of 0.83, 0.62, 0.76, and 0.57, respectively.

The nomogram was constructed using the same parameters selected for the prediction model. It demonstrated good predictive performance on the calibration curve.

Overall, our findings support the significance of LAR in AMI-CS patients, showing a strong statistical association and consistency with previous research. However, external validation in larger, multicenter cohorts is needed before clinical application.

5. Limitations

Although the model performed well, it has some limitations. The sample size was very small, and with only 23 in-hospital mortalities, there is a lower events-per-variable ratio, which increases the risk of multicollinearity and overfitting. Another limitation of this study is that subgroup analysis between STEMI and non-STEMI patients was not performed because of the limited sample size. Future studies with larger sample sizes are needed to evaluate the predictive value of LAR in these 2 patient groups. Also, detailed data on reperfusion strategies, such as primary PCI, salvage PCI, CABG, and thrombolytic therapy, were not available. In addition, data on infarction location, culprit vessel, and the extent of coronary artery disease were also not available. LASSO reduced this problem to some extent, but the estimates may still be unstable, so results should be interpreted cautiously. Excluding patients with missing laboratory results and those who died at the very beginning of CS may have introduced selection bias, as the most severe cases were left out, and this could reduce the applicability of our findings. LAR was only measured once at admission, not dynamically, so we could not evaluate whether changes over time add prognostic value. Society for Cardiovascular Angiography and Interventions stages were not recorded in our study, and therefore could not be analyzed or reported. This may have affected the assessment of shock severity in our study. Because the study was conducted in a single center, external validation in other hospitals is still required before the model can be widely applied. A formal screening log was not kept, which affects transparency regarding excluded cases. Maintaining a record of screened and excluded patients, along with serial biomarker measurement, would improve clarity and reproducibility in future research. Thus, with proper consideration of these limitations, further studies can eventually contribute to the development of more effective prediction models in clinical practice.

6. Conclusion

The nomogram created from the prediction model study using LAR, cold and clammy skin, AST, CRP, and LVEF < 40% may show a good predictive outcome in AMI-CS patients, with good prediction of in-hospital mortality. Compared with the previously available risk scoring systems (APACHE II and CardShock), it might demonstrate superior results. LAR, used as a new biomarker in the prediction model, also exhibits significant prognostic value in predicting in-hospital mortality in AMI-CS patients.

Acknowledgments

We express our gratitude to the colleagues from the CCU team of Renmin Hospital, Hubei University of Medicine, for assisting in the collection of case data. We also thank all the patients enrolled in this study for their participation.

Author contributions

Data curation: Roshan Angdembe, You Wu, Qun Ke.

Formal analysis: Roshan Angdembe, You Wu, Qun Ke, Yinsheng Jin.

Writing – original draft: Roshan Angdembe.

Writing – review & editing: Lingling Yao.