Albumin-based nutritional-inflammatory indices and prognosis in critically ill patients with acute myocardial infarction: a MIMIC-IV cohort study

Abstract

Objective

This study investigated the associations between 12 albumin-based nutritional–inflammatory indices and adverse outcomes in critically ill patients with acute myocardial infarction and compared their prognostic performance.

Methods

A retrospective cohort study was conducted using the MIMIC-IV database, which includes data from 2,255 critically ill AMI patients. Twelve nutritional and inflammatory indices were assessed: Advanced Lung Cancer Inflammation Index (ALI), Blood Urea Nitrogen to Albumin Ratio (BAR), Creatinine to Albumin Ratio (CAR), Controlling Nutritional Status (CONUT) score, Geriatric Nutritional Risk Index (GNRI), Hemoglobin-Albumin-Lymphocyte-Platelet (HALP) score, Lactate to Albumin Ratio (LAR), Monocyte to Albumin Ratio (MAR), Neutrophil to Albumin Ratio (NAR), Neutrophil Percentage to Albumin Ratio (NPAR), Prognostic Nutritional Index (PNI), and Red Cell Distribution Width to Albumin Ratio (RAR). The primary outcomes were 30-day and 365-day all-cause mortality (ACM), while secondary outcomes included new-onset atrial fibrillation (NOAF), new-onset ventricular arrhythmia (NOVA), and acute heart failure (AHF).

Results

In fully adjusted models, lower levels of GNRI, HALP, and PNI, and higher levels of BAR, CAR, CONUT, LAR, NPAR, and RAR were independently associated with increased 30-day and 365-day ACM. Restricted cubic spline analyses revealed nonlinear relationships for several indices (ALI, BAR, CAR, LAR, PNI). Among the indices evaluated, LAR showed the highest discriminatory performance for 30-day all-cause mortality (AUC = 0.719), whereas BAR showed the highest discriminatory performance for 365-day all-cause mortality (AUC = 0.717). Incremental predictive analyses further supported the additional prognostic information provided by these indices compared with albumin alone. Multiple nutritional indices were significantly associated with both NOAF and NOVA during AMI hospitalization, whereas no significant associations were found with AHF or mechanical complications. However, the findings for mechanical complications should be interpreted cautiously because of the limited number of events.

Conclusion

Albumin-based nutritional–inflammatory indices provide additional prognostic information in critically ill patients with acute myocardial infarction. LAR and BAR showed the greatest discriminatory performance for short- and long-term mortality, respectively, but should be considered adjunctive risk-stratification markers rather than stand-alone clinical decision tools.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12872-026-06613-y.

Introduction

With the aging of the population and changes in lifestyle and diet, the incidence of acute myocardial infarction (AMI) is on the rise globally [1, 2]. AMI has an acute onset and rapid progression, posing a serious threat to public health and being one of the leading causes of death globally, imposing a huge burden on the socioeconomy [3]. Although early percutaneous coronary intervention (PCI) and optimized pharmacological therapy have been widely applied in AMI patients and have significantly improved the prognosis [4], the incidence of major adverse cardiovascular events remains high, especially in critically ill patients [5]. Therefore, it is particularly important to select appropriate early risk stratification methods and identify risk factors leading to adverse outcomes for critically ill AMI patients.

Recent studies have indicated that inflammation, as a systemic and dynamic pathological process, exerts various adverse effects on the short-term progression and long-term prognosis of AMI patients, including myocardial injury, promotion of atherosclerosis, and thrombosis [6, 7]. In addition, the prognosis of AMI patients is not only associated with the level of inflammation in the body but also closely related to nutritional status, even in obese patients [8–10]. Some scholars have even proposed that in atherosclerosis, inflammation and malnutrition interact and promote each other [11]. However, previous research has suggested that the prognostic predictive ability of single indicators such as serum albumin and CRP in AMI patients is still limited [12]. A variety of composite albumin-based indicators have been identified to further enhance the accuracy. In recent years, albumin-related indices that integrate inflammation and nutrition, such as the C-Reactive Protein to Albumin Ratio, Blood Urea Nitrogen to Albumin Ratio (BAR), Prognostic Nutritional Index (PNI), and Geriatric Nutritional Risk Index (GNRI), have garnered increasing attention. These indices have been proven to be associated with the occurrence and development of various diseases and have been preliminarily validated in cardiovascular diseases. For example, the C-Reactive Protein to Albumin Ratio, as a reliable and easily accessible tool for dynamic prognostic monitoring, integrates the patient’s pro-inflammatory and nutritional status and has been shown to be useful for risk stratification in patients with valvular heart disease (VHD) [13]. Yu-Lun Cheng et al. also demonstrated that the Prognostic Nutritional Index (PNI), derived from serum albumin levels and total lymphocyte count, is independently associated with long-term all-cause and cardiovascular mortality in patients with acute heart failure [14]. Additionally, the BAR can also serve as a composite marker integrating nitrogen metabolism and protein reserve, dynamically reflecting the comprehensive state of nutritional depletion and inflammatory activation. It has significant incremental value in risk stratification and prognostic assessment in heart failure, critical illness, and infectious diseases [15].

These indices have all demonstrated good prognostic value in patients with cardiovascular diseases such as AMI. However, there is currently a lack of systematic comparison of the performance of these indices in adverse outcomes among critically ill AMI patients, and many albumin-based nutritional and inflammatory indices remain underexplored. This study aims to systematically compare the performance of different nutritional and inflammatory indices in adverse outcomes among critically ill AMI patients based on large-scale inpatient data, in order to provide evidence-based support for precise risk assessment and individualized interventions. The specific nutritional indices include the Advanced Lung Cancer Inflammation Index (ALI), BAR, Creatinine to Albumin Ratio (CAR), Controlling Nutritional Status (CONUT) score, GNRI, Hemoglobin-Albumin-Lymphocyte-Platelet (HALP) score, Lactate to Albumin Ratio (LAR), Monocyte to Albumin Ratio (MAR), Neutrophil to Albumin Ratio (NAR), Neutrophil Percentage to Albumin Ratio (NPAR), PNI, and Red Cell Distribution Width to Albumin Ratio (RAR).

Methods

Study population

The dataset utilized in this study was sourced from the MIMIC-IV database (version 3.1), a publicly accessible and de-identified repository of medical records. It documents the hospitalizations of patients at the Beth Israel Deaconess Medical Center from 2008 to 2022, covering 65,366 intensive care unit (ICU) admissions. Access to the database was granted to the author (HW) (Record ID: 63644491). Given the de-identified nature of the data, patient consent was not required. The inclusion and exclusion criteria for the study population are as follows: (1) Patients who experienced acute myocardial infarction (AMI) during hospitalization; (2) Only the first ICU admission information was extracted for each patient; (3) Patients with missing data on nutritional and inflammatory indices were excluded; (4) Patients with missing low-density lipoprotein (LDL) data were excluded (Supplementary Fig. 1).

Data collection and definitions

Structured Query Language (SQL) and PostgreSQL (version 16.0) were utilized for data extraction to collect the basic characteristics of the study subjects, which included demographic data, severity scores, laboratory indicators, comorbidities, and medication use. Demographic data, laboratory indicators, and vital signs were obtained from the initial measurements after admission. For variables with a missing rate below 20%, random forest imputation was employed to fill in the missing values. The diagnoses of chronic kidney disease (CKD), hypertension (HTN), ischemic stroke, hyperlipidemia, chronic obstructive pulmonary disease (COPD), diabetes mellitus (DM), severe liver disease, gastrointestinal bleeding (GI), cardiogenic shock (CS), acute heart failure (AHF), and mechanical complications (mitral valve prolapse or rupture, ventricular septal defect, and cardiac rupture) were determined according to the International Classification of Diseases (ICD-9 and ICD-10) classifications. The ICD diagnostic codes for AMI are provided in Supplementary Table 1. Rhythm data were extracted from the MIMIC-IV “icu.chartevents” table and were defined based on cardiac rhythm information documented by bedside nurses. Rhythm-related “itemid” and “value” fields were used to identify atrial fibrillation, ventricular tachycardia, ventricular flutter, and ventricular fibrillation. Records were linked to the “icu.icustays” table using “stay_id”, and event timing was determined by comparing “chartevents.charttime” with “icustays.intime”. After excluding patients with evidence of the corresponding arrhythmia before ICU admission or within the first 24 h after admission, New-onset atrial fibrillation (NOAF), new-onset ventricular arrhythmia (NOVA) were defined as the first qualifying rhythm event documented more than 24 h after ICU admission. NOVA included ventricular tachycardia, ventricular flutter, and ventricular fibrillation [16]. In contrast, AHF and mechanical complications were identified from diagnostic records during the index hospitalization; the exact timing of diagnosis could not be clearly determined, and they were not defined as strictly new‑onset events.Medications were identified using the “gsn” coding in the “prescriptions” table of the MIMIC-IV database. The calculation formulas for nutritional and inflammatory indicators, including ALI, BAR, CAR, CONUT, GNRI, HALP, LAR, MAR, NAR, NPAR, PNI, and RAR, are detailed in Supplementary Table 2.

Clinical outcomes

The primary outcomes were 30‑day and 365‑day all‑cause mortality (ACM). Secondary outcomes included NOAF, NOVA, AHF, and mechanical complications recorded during the index hospitalization. The primary outcomes clearly occurred after the measurement of the exposure indices. The exact timing of the secondary outcomes relative to the first exposure index measurement could not be uniformly determined; therefore, these analyses were used only to explore associations between documented in‑hospital complications and nutritional‑inflammatory indices.

Statistical analysis

Continuous variables were assessed for normality using the Shapiro–Wilk test and visual inspection of histograms and quantile–quantile plots. As most continuous variables were non-normally distributed, they were presented as medians with interquartile ranges and compared between independent groups using the Wilcoxon rank-sum test (Mann–Whitney U test). Categorical variables were summarized as counts and percentages and compared using the chi-square test or Fisher’s exact test, as appropriate. Kaplan-Meier (K-M) curves were utilized to analyze the incidence of the primary clinical endpoint across different levels of each of the 12 nutritional and inflammatory indices. Confounding variables were identified using LASSO regression based on the regularized L1 method. Multivariate Cox regression analysis was employed to elucidate the correlation between nutritional and inflammatory indices and the primary endpoint, with three models constructed based on confounding variables extracted by LASSO regression (Supplementary Fig. 2). Model 1 was unadjusted, Model 2 adjusted for age, gender, smoking, alcohol and BMI, while Model 3 adjusted for age, gender, smoking, alcohol, BMI, temperature, respiratory rate (RR), SpO₂, platelets (PLT), lactate, cardiac troponin T (cTnT), low-density lipoprotein (LDL), glucose (Glu), lactate dehydrogenase (LDH), potassium, stroke, hypertension (HTN), angiotensin-converting enzyme inhibitors/angiotensin receptor blockers (ACEI/ARB), beta-blockers, statins, and non-vitamin K antagonist oral anticoagulants (NOAC). Multicollinearity was assessed using variance inflation factors, and the proportional hazards assumption was examined with Schoenfeld residuals (Supplementary Table S3 and Supplementary Table S4). Restricted cubic spline and two-piecewise regression analyses were used to explore potential nonlinear associations and estimate data-derived inflection points. These analyses were considered exploratory rather than confirmatory. In addition, a model including albumin alone was used as the reference to evaluate the incremental predictive value of the 12 nutritional–inflammatory indices for the primary outcome. Differences in AUC were assessed using the DeLong test, and ΔAUC, continuous net reclassification improvement (NRI), and integrated discrimination improvement (IDI) were reported. The two indices with the greatest incremental predictive value were then selected for subgroup analyses to determine whether their associations with the primary clinical outcome varied by age (≤ 65 vs. > 65 years), BMI (≤ 28 vs. > 28 kg/m²), gender, smoking status, CKD, HTN, and DM. Interaction P values were calculated by including multiplicative interaction terms in the corresponding models. Results were presented in forest plots. Exploratory univariable and multivariable logistic regression analyses were performed for secondary outcomes to evaluate the associations between nutritional‑inflammatory indices and complications recorded during hospitalization. These results should not be interpreted as time‑to‑event relationships, prospective predictive effects, or causal associations. To assess the robustness of the findings, the following sensitivity analyses were conducted: missing LDL values were handled using multiple imputation, and the Cox regression analyses for the primary outcomes were repeated; to avoid potential overadjustment arising from the simultaneous inclusion of nutritional–inflammatory indices and their constituent variables, BMI, platelet count, or lactate was excluded from Model 3, as appropriate; percutaneous coronary intervention, cardiogenic shock, vasopressor use, and mechanical ventilation were additionally adjusted for based on Model 3; and the SOFA score was further included in the fully adjusted model to reassess the associations between nutritional–inflammatory indices and 30‑day and 365‑day all‑cause mortality. Statistical analyses were performed using R software (version 4.5.1), with a significance level of P < 0.05 for two-sided tests.

Results

Baseline characteristics

This study enrolled 2,255 critically ill patients, mostly male, with a median age of 71. Of these, 565 and 892 patients died by 30 days and 1 year, respectively (Table 1 and Supplementary Table 5). In both 30- and 365-day mortality groups, patients who died shared similar demographic and clinical characteristics. They were mostly older adults, female, smokers, and had a lower BMI (Table 1 and S2). At admission, they had higher heart and respiratory rates, lower oxygen saturation and blood pressure. Their first measured WBC, neutrophils, RDW, LDH, lactate, creatinine, BUN, AST, ALT, LDL, glucose, potassium, cTnT, and CKMB were significantly higher, while albumin and hemoglobin were significantly lower. They also had higher prevalence of CKD, stroke, severe liver disease, GI bleeding, NOVA, and NOAF. Moreover, they less frequently used aspirin, ACEI/ARB, β-blockers, statins, and NOAC. Among 12 albumin-based nutritional indicators, their ALI, GNRI, HALP, and PNI were lower, whereas BAR, CAR, CONUT, LAR, MAR, NAR, NPAR, and RAR were higher (Table 1 and Supplementary Table 5).

Table 1.

Baseline characteristics of patients with AMI according to 30-day survival status

Variables	Overall(2255)	30 days All-cause mortality		P-value
		Alive (N = 1690)	Dead (N = 565)
Age, years	71.00 [61.50, 80.00]	70.00 [60.00, 78.00]	76.00 [66.00, 84.00]	< 0.001
Gender (%)				0.01
Female	815 (36.1)	585 (34.6)	230 (40.7)
Male	1440 (63.9)	1105 (65.4)	335 (59.3)
Height, cm	170.00 [163.00, 178.00]	170.00 [163.00, 178.00]	168.00 [160.00, 175.00]	0.005
Weight, kg	79.00 [66.95, 93.20]	79.90 [68.00, 93.80]	76.00 [61.80, 91.00]	< 0.001
BMI, kg/m2	27.28 [23.92, 31.78]	27.63 [24.19, 31.86]	26.23 [22.96, 31.74]	< 0.001
Alcohol (%)				0.701
No	2028 (89.9)	1517 (89.8)	511 (90.4)
Yes	227 (10.1)	173 (10.2)	54 ( 9.6)
Smoking (%)				< 0.001
No	1211 (53.7)	1122 (66.4)	89 (15.8)
Yes	1044 (46.3)	568 (33.6)	476 (84.2)
GCS	15.00 [15.00, 15.00]	15.00 [15.00, 15.00]	15.00 [14.00, 15.00]	< 0.001
SOFA	1.00 [0.00, 4.00]	1.00 [0.00, 3.00]	2.00 [0.00, 4.00]	< 0.001
Vital signs
Temperature, ℃	36.70 [36.40, 37.00]	36.70 [36.50, 37.00]	36.60 [36.30, 37.00]	< 0.001
HR, bpm	87.00 [76.00, 101.00]	85.00 [75.00, 99.00]	93.00 [78.00, 107.00]	< 0.001
RR, insp/min	20.00 [16.00, 24.00]	19.00 [15.00, 23.00]	21.00 [17.00, 25.00]	< 0.001
SpO2, %	98.00 [95.00, 100.00]	98.00 [95.00, 100.00]	97.00 [94.00, 100.00]	< 0.001
SBP, mmhg	118.00 [103.00, 135.00]	119.00 [105.00, 136.00]	115.00 [99.00, 133.00]	< 0.001
DBP, mmhg	68.00 [58.00, 81.00]	69.00 [59.00, 81.00]	66.00 [56.00, 79.00]	0.006
Laboratory indicators
RBC, m/uL	4.12 [3.51, 4.63]	4.18 [3.58, 4.67]	3.92 [3.27, 4.47]	< 0.001
WBC, k/uL	10.40 [7.50, 14.80]	10.10 [7.40, 14.00]	11.30 [7.80, 17.20]	< 0.001
PLT, k/uL	217.00 [168.00, 274.00]	219.00 [173.00, 273.00]	208.00 [152.00, 277.00]	0.012
Hematocrit, %	37.60 [32.40, 41.85]	38.00 [33.10, 42.20]	36.20 [30.70, 40.70]	< 0.001
RDW, %	13.90 [13.20, 15.25]	13.80 [13.10, 15.00]	14.50 [13.40, 16.30]	< 0.001
Lymphocytes, 103/µL	1.36 [0.82, 1.97]	1.40 [0.87, 2.01]	1.17 [0.70, 1.76]	< 0.001
Monocytes, 103/µL	0.62 [0.42, 0.94]	0.62 [0.43, 0.93]	0.62 [0.38, 0.99]	0.472
Neutrophils, 103/µL	7.75 [4.92, 12.03]	7.40 [4.83, 11.40]	8.99 [5.55, 14.75]	< 0.001
Hemoglobin, g/dL	12.30 [10.40, 13.90]	12.50 [10.70, 14.00]	11.70 [9.90, 13.40]	< 0.001
Albumin, g/dL	3.50 [3.00, 3.90]	3.60 [3.10, 4.00]	3.20 [2.60, 3.70]	< 0.001
Lactate, mmol/L	1.80 [1.20, 3.00]	1.60 [1.20, 2.50]	2.50 [1.70, 5.10]	< 0.001
TBIL, mg/dL	0.60 [0.40, 0.90]	0.60 [0.40, 0.80]	0.70 [0.40, 1.20]	< 0.001
AST, IU/L	53.00 [27.00, 140.00]	45.00 [25.00, 115.00]	83.00 [39.00, 249.00]	< 0.001
ALT, IU/L	31.00 [18.00, 70.00]	28.50 [17.00, 56.00]	48.00 [21.00, 126.00]	< 0.001
Cre, mg/dL	1.20 [0.90, 1.90]	1.10 [0.90, 1.70]	1.60 [1.10, 2.40]	< 0.001
BUN, mg/dL	25.00 [17.00, 42.50]	22.00 [16.00, 37.00]	34.00 [22.00, 54.00]	< 0.001
Tg, mg/dL	118.00 [85.00, 165.00]	118.00 [84.00, 165.00]	120.00 [87.00, 166.00]	0.297
Tc, mg/dL	149.00 [116.00, 186.00]	139.00 [103.00, 176.00]	152.50 [121.00, 187.00]	< 0.001
HDL, mg/dL	43.00 [33.00, 54.00]	43.00 [34.00, 53.00]	43.00 [30.00, 55.00]	0.182
LDL, mg/dL	77.00 [56.00, 108.50]	72.00 [50.00, 100.00]	80.00 [59.00, 111.00]	< 0.001
cTnT, ng/mL	0.22 [0.07, 0.80]	0.20 [0.07, 0.74]	0.26 [0.08, 0.95]	0.037
CKMB, IU/dL	7.00 [3.00, 23.00]	6.00 [3.00, 22.00]	9.00 [4.00, 25.00]	0.001
Glu, mg/dL	138.00 [109.00, 195.00]	135.00 [107.00, 185.00]	151.00 [115.00, 222.00]	< 0.001
HbA1c, %	5.90 [5.50, 7.10]	5.90 [5.50, 7.30]	5.80 [5.40, 6.90]	0.002
Sodium, mmol/L	138.00 [136.00, 141.00]	139.00 [136.00, 141.00]	138.00 [135.00, 141.00]	0.599
Potassium, mmol/L	4.30 [3.90, 4.80]	4.30 [3.90, 4.70]	4.50 [4.00, 5.10]	< 0.001
LDH, mg/dL	348.00 [243.00, 587.50]	314.00 [229.25, 488.00]	509.00 [316.00, 910.00]	< 0.001
Comorbidities, n (%)
CKD				< 0.001
No	1515 (67.2)	1170 (69.2)	345 (61.1)
Yes	740 (32.8)	520 (30.8)	220 (38.9)
HTN				0.928
No	532 (23.6)	400 (23.7)	132 (23.4)
Yes	1723 (76.4)	1290 (76.3)	433 (76.6)
Stroke				< 0.001
No	2046 (90.7)	1560 (92.3)	486 (86.0)
Yes	209 ( 9.3)	130 ( 7.7)	79 (14.0)
Hyperlipidemia				0.596
No	1018 (45.1)	757 (44.79)	261 (46.19)
Yes	1237 (54.9)	933 (55.21)	304 (53.81)
COPD				0.282
No	1630 (72.3)	1232 (72.9)	398 (70.4)
Yes	625 (27.7)	458 (27.1)	167 (29.6)
DM				0.193
No	1266 (56.1)	935 (55.3)	331 (58.6)
Yes	989 (43.9)	755 (44.7)	234 (41.4)
Severe_liver_disease				< 0.001
No	2169 (96.2)	1641 (97.1)	528 (93.5)
Yes	86 ( 3.8)	49 ( 2.9)	37 ( 6.5)
GI				0.018
No	2147 (95.2)	1620 (95.9)	527 (93.3)
Yes	108 ( 4.8)	70 ( 4.1)	38 ( 6.7)
CS
No	1779 (78.89)	1416 (83.79)	363 (64.25)	< 0.001
Yes	476 (21.11)	274 (16.21)	202 (35.75)
VA				< 0.001
No	2160 (95.8)	1639 (97.0)	521 (92.2)
Yes	95 ( 4.2)	51 ( 3.0)	44 ( 7.8)
AF				< 0.001
No	1950 (86.5)	1501 (88.8)	449 (79.5)
Yes	305 (13.5)	189 (11.2)	116 (20.5)
AHF				0.057
No	1905 (84.5)	1413 (83.6)	492 (87.1)
Yes	350 (15.5)	277 (16.4)	73 (12.9)
Mechanical_Complications				0.913
No	2230 (98.9)	1672 (98.9)	558 (98.8)
Yes	25 ( 1.1)	18 ( 1.1)	7 ( 1.2)
Drugs and treatments, n (%)
Aspirin				< 0.001
No	428 (19.0)	262 (15.5)	166 (29.4)
Yes	1827 (81.0)	1428 (84.5)	399 (70.6)
ACEI/ARB				< 0.001
No	1305 (57.9)	813 (48.1)	492 (87.1)
Yes	950 (42.1)	877 (51.9)	73 (12.9)
Betablocker				< 0.001
No	579 (25.7)	270 (16.0)	309 (54.7)
Yes	1676 (74.3)	1420 (84.0)	256 (45.3)
Statin				< 0.001
No	482 (21.4)	264 (15.6)	218 (38.6)
Yes	1773 (78.6)	1426 (84.4)	347 (61.4)
NOAC				< 0.001
No	1936 (85.9)	1405 (83.1)	531 (94.0)
Yes	319 (14.1)	285 (16.9)	34 ( 6.0)
Vasopressors				< 0.001
No	1083 (48.03)	945 (55.92)	138 (24.42)
Yes	1172 (51.97)	745 (44.08)	427 (75.58)
PCI				0.061
No	172 (7.63)	143 (8.46)	29 (5.13)
Yes	2083 (92.37)	1547 (91.54)	536 (94.87)
Mechanical ventilation				< 0.001
No	1204 (53.39)	969 (57.34)	235 (41.59)
Yes	1051 (46.61)	721 (42.66)	330 (58.41)
Nutritional indicators
ALI	15.49 [7.09, 35.27]	17.64 [8.21, 38.10]	10.43 [4.55, 25.41]	< 0.001
BAR	7.35 [4.55, 13.33]	6.41 [4.15, 11.48]	11.36 [6.92, 18.26]	< 0.001
CAR	3.55 [2.43, 6.00]	3.21 [2.31, 5.17]	5.14 [3.41, 8.18]	< 0.001
CONUT	4.00 [2.00, 6.00]	4.00 [2.00, 6.00]	6.00 [3.00, 8.00]	< 0.001
GNRI	93.82 [85.21, 99.77]	94.22 [87.40, 99.77]	87.86 [80.41, 95.30]	< 0.001
HALP	24.68 [13.98, 36.95]	26.39 [15.18, 38.52]	19.70 [10.14, 31.44]	< 0.001
LAR	0.53 [0.36, 0.91]	0.47 [0.33, 0.74]	0.85 [0.51, 1.65]	< 0.001
MAR	0.18 [0.12, 0.28]	0.18 [0.12, 0.27]	0.20 [0.12, 0.32]	0.011
NAR	2.30 [1.37, 3.75]	2.17 [1.32, 3.41]	2.88 [1.68, 4.87]	< 0.001
NPAR	22.01 [17.79, 26.67]	21.26 [17.27, 25.63]	24.50 [20.18, 30.25]	< 0.001
PNI	42.05 [36.08, 47.70]	42.95 [37.35, 48.45]	38.65 [32.35, 44.21]	< 0.001
RAR	4.10 [3.49, 5.04]	3.92 [3.41, 4.74]	4.69 [3.88, 5.84]	< 0.001

BMI Body mass index, GCS Glasgow Coma Scale, SOFA Sequential Organ Failure Assessment, HR Heart rate, RR Respiratory rate, SpO2 oxyhemoglobin saturation, SBP Systolic blood pressure, DBP Diastolic blood pressure, RBC Red blood cell, WBC White blood cell, PLT Platelet, RDW Red cell distribution width, TBIL Total bilirubin, AST Aspartate aminotransferase, ALT Alanine aminotransferase, Cre Creatinine, BUN Blood urea nitrogen, Tg Triglyceride, Tc Total cholesterol, HDL High-density lipoprotein, LDL Low-density lipoprotein, cTnT Cardiac troponin T, CKMB Creatine kinase-MB, Glu Glucose, HbA1c Glycated hemoglobin, CKD Chronic kidney disease, HTN Hypertension, COPD Chronic obstructive pulmonary disease, DM Diabetes mellitus, GI Gastrointestinal, CS Cardiogenic shock, VA Ventricular arrhythmia, AF Atrial fibrillation, AHF Acute heart failure, ACEI/ARB Angiotensin-converting enzyme inhibitors/angiotensin receptor blockers, NOAC Novel oral anticoagulants

Survival analysis

Using K-M survival analysis, the 12 nutritional indicators were divided into four groups by quartiles to compare the ACM rates of patients grouped at different levels. The analysis found that patients with lower levels of ALI, GNRI, HALP, and PNI had a higher risk of 30-day mortality; while patients with higher levels of BAR, CAR, CONUT, LAR, MAR, NAR, NPAR, and RAR had a higher mortality rate (Fig. 1). In addition, there was a significant difference in ACM between the groups (log-rank test P-value < 0.001). Similarly, consistent results were observed in 365-day mortality risk (Supplementary Fig. 3).

Fig. 1.

Kaplan-Meier survival analysis curve for 30-day ACM

The associations between primary clinical endpoints and the nutritional indicators

To better assess the precise impact of various nutritional indicators on ACM in critically ill AMI patients, three Cox proportional hazards models were employed. In the unadjusted model, all nutritional indicators were significantly associated with 30-day ACM in AMI patients (Table 2). In the fully adjusted Model 3, ALI was significant only in Group 4, MAR was not statistically significant, and NAR no longer had a simple stepwise increasing relationship with mortality risk. Specifically, lower levels of GNRI (0.80 [0.73–0.87], P < 0.001), HALP (0.87 [0.80–0.95], P = 0.002), and PNI (0.80 [0.65–0.99], P = 0.042) were associated with higher ACM in critically ill AMI patients, while higher levels of BAR (1.15 [1.08–1.23], P = 0.042), CAR (1.13 [1.05–1.22], P = 0.001), CONUT (1.20 [1.10–1.31], P < 0.001), LAR (1.24 [1.18–1.30], P < 0.001), NPAR (1.18 [1.10–1.26], P < 0.001), and RAR (1.16 [1.09–1.24], P = 0.001) were associated with higher mortality risk (Table 2). Consistent results were observed for 365-day ACM (Supplementary Table 6).

Table 2.

Cox proportional hazards models for 30 days all-cause mortality

Categories	Model1		Model2		Model3
	HR (95% CI)	P-value	HR (95% CI)	P-value	HR (95% CI)	P-value
ALI
Q1	Ref.		Ref.		Ref.
Q2	0.63 (0.51–0.77)	P < 0.001	0.65 (0.53–0.80)	P = 0.021	0.93 (0.75–1.15)	P = 0.510
Q3	0.42 (0.34–0.54)	P < 0.001	0.46 (0.36–0.58)	P = 0.001	0.82 (0.64–1.05)	P = 0.113
Q4	0.43 (0.34–0.54)	P < 0.001	0.48 (0.38–0.61)	P < 0.001	0.77 (0.60–0.98)	P = 0.034
Per SD	0.19 (0.05–0.76)	P = 0.020	0.33 (0.10–1.12)	P = 0.076	0.64 (0.30–1.35)	P = 0.238
BAR
Q1	Ref.		Ref.		Ref.
Q2	2.62 (1.84–3.74)	P < 0.001	2.43 (1.70–3.47)	P < 0.001	1.76 (1.22–2.52)	P = 0.002
Q3	4.96 (3.56–6.92)	P < 0.001	4.33 (3.09–6.06)	P < 0.001	2.57 (1.82–3.62)	P < 0.001
Q4	6.59 (4.76–9.12)	P < 0.001	5.90 (4.25–8.20)	P < 0.001	2.63 (1.86–3.71)	P < 0.001
Per SD	1.39 (1.31–1.46)	P < 0.001	1.37 (1.30–1.45)	P < 0.001	1.15 (1.08–1.23)	P = 0.042
CAR
Q1	Ref.		Ref.		Ref.
Q2	2.41 (1.73–3.37)	P < 0.001	2.30 (1.65–3.22)	P < 0.001	2.03 (1.44–2.85)	P < 0.001
Q3	3.98 (2.90–5.45)	P < 0.001	3.64 (2.65-5.00)	P < 0.001	2.48 (1.79–3.43)	P < 0.001
Q4	5.76 (4.24–7.82)	P < 0.001	5.64 (4.15–7.67)	P < 0.001	2.56 (1.85–3.54)	P < 0.001
Per SD	1.19 (1.13–1.25)	P < 0.001	1.24 (1.18–1.31)	P < 0.001	1.13 (1.05–1.22)	P = 0.001
CONUT
Q1	Ref.		Ref.		Ref.
Q2	1.63 (1.24–2.14)	P < 0.001	1.51 (1.15–1.99)	P = 0.003	1.22 (0.92–1.61)	P = 0.174
Q3	2.36 (1.81–3.08)	P < 0.001	2.11 (1.61–2.75)	P < 0.001	1.52 (1.15–2.01)	P = 0.003
Q4	3.48 (2.74–4.44)	P < 0.001	3.10 (2.43–3.95)	P < 0.001	1.64 (1.26–2.14)	P < 0.001
Per SD	1.60 (1.48–1.73)	P < 0.001	1.55 (1.43–1.68)	P < 0.001	1.20 (1.10–1.31)	P < 0.001
GNRI
Q1	Ref.		Ref.		Ref.
Q2	0.58 (0.48–0.71)	P < 0.001	0.58 (0.47–0.70)	P < 0.001	0.86 (0.70–1.06)	P = 0.152
Q3	0.41 (0.32–0.52)	P < 0.001	0.42 (0.33–0.53)	P < 0.001	0.71 (0.56–0.91)	P = 0.006
Q4	0.32 (0.24–0.41)	P < 0.001	0.35 (0.27–0.46)	P < 0.001	0.65 (0.49–0.86)	P = 0.003
Per SD	0.62 (0.58–0.67)	P < 0.001	0.63 (0.59–0.69)	P < 0.001	0.80 (0.73–0.87)	P < 0.001
HALP
Q1	Ref.		Ref.		Ref.
Q2	0.82 (0.67–1.02)	P = 0.071	0.83 (0.67–1.03)	P = 0.086	0.99 (0.80–1.23)	P = 0.944
Q3	0.61 (0.48–0.76)	P < 0.001	0.67 (0.53–0.84)	P < 0.001	0.79 (0.62-1.00)	P = 0.046
Q4	0.48 (0.38–0.61)	P < 0.001	0.56 (0.43–0.71)	P < 0.001	0.74 (0.58–0.96)	P = 0.020
Per SD	0.73 (0.67–0.80)	P < 0.001	0.77 (0.71–0.84)	P < 0.001	0.87 (0.80–0.95)	P = 0.002
LAR
Q1	Ref.		Ref.		Ref.
Q2	1.72 (1.24–2.39)	P = 0.001	1.63 (1.17–2.28)	P = 0.004	1.23 (0.88–1.72)	P = 0.220
Q3	2.96 (2.18–4.03)	P < 0.001	2.84 (2.09–3.87)	P < 0.001	1.85 (1.35–2.54)	P < 0.001
Q4	6.35 (4.76–8.47)	P < 0.001	5.86 (4.39–7.83)	P < 0.001	2.25 (1.60–3.15)	P < 0.001
Per SD	1.41 (1.35–1.46)	P < 0.001	1.43 (1.37–1.49)	P < 0.001	1.24 (1.18–1.30)	P < 0.001
MAR
Q1	Ref.		Ref.		Ref.
Q2	0.72 (0.56–0.92)	P = 0.010	0.70 (0.54–0.89)	P = 0.004	0.74 (0.58–0.95)	P = 0.017
Q3	0.88 (0.70–1.12)	P = 0.301	0.81 (0.64–1.03)	P = 0.080	0.82 (0.64–1.04)	P = 0.095
Q4	1.25 (1.01–1.56)	P = 0.043	1.16 (0.93–1.44)	P = 0.191	0.82 (0.65–1.02)	P = 0.075
Per SD	1.07 (1.03–1.11)	P < 0.001	1.06 (1.02–1.10)	P = 0.002	1.00 (0.95–1.06)	P = 0.875
NAR
Q1	Ref.		Ref.		Ref.
Q2	1.21 (0.93–1.58)	P = 0.154	1.18 (0.91–1.54)	P = 0.214	1.32 (1.01–1.73)	P = 0.043
Q3	1.33 (1.02–1.73)	P = 0.032	1.27 (0.98–1.65)	P = 0.074	1.26 (0.96–1.64)	P = 0.090
Q4	2.51 (1.98–3.19)	P < 0.001	2.49 (1.97–3.16)	P < 0.001	1.49 (1.16–1.91)	P = 0.002
Per SD	1.18 (1.14–1.22)	P < 0.001	1.15 (1.11–1.19)	P < 0.001	1.03 (0.98–1.07)	P = 0.245
NPAR
Q1	Ref.		Ref.		Ref.
Q2	1.07 (0.80–1.42)	P = 0.663	1.00 (0.75–1.34)	P = 0.974	1.05 (0.79–1.41)	P = 0.731
Q3	1.89 (1.46–2.45)	P < 0.001	1.75 (1.35–2.27)	P < 0.001	1.44 (1.10–1.87)	P = 0.007
Q4	2.89 (2.26–3.69)	P < 0.001	2.63 (2.06–3.37)	P < 0.001	1.52 (1.18–1.96)	P < 0.001
Per SD	1.30 (1.24–1.36)	P < 0.001	1.31 (1.24–1.37)	P < 0.001	1.18 (1.10–1.26)	P < 0.001
PNI
Q1	Ref.		Ref.		Ref.
Q2	0.68 (0.55–0.83)	P < 0.001	0.66 (0.54–0.81)	P < 0.001	0.94 (0.76–1.16)	P = 0.576
Q3	0.42 (0.33–0.53)	P < 0.001	0.43 (0.34–0.55)	P < 0.001	0.78 (0.61–0.99)	P = 0.039
Q4	0.35 (0.28–0.45)	P < 0.001	0.40 (0.31–0.52)	P < 0.001	0.67 (0.52–0.87)	P = 0.003
Per SD	0.39 (0.30–0.52)	P < 0.001	0.45 (0.34–0.60)	P < 0.001	0.80 (0.65–0.99)	P = 0.042
RAR
Q1	Ref.		Ref.		Ref.
Q2	1.54 (1.14–2.06)	P = 0.004	1.39 (1.03–1.87)	P = 0.031	1.13 (0.83–1.53)	P = 0.446
Q3	2.11 (1.59–2.79)	P < 0.001	1.87 (1.41–2.48)	P < 0.001	1.37 (1.03–1.82)	P = 0.032
Q4	3.88 (2.99–5.04)	P < 0.001	3.50 (2.69–4.56)	P < 0.001	1.67 (1.27–2.20)	P < 0.001
Per SD	1.27 (1.22–1.31)	P < 0.001	1.28 (1.23–1.33)	P < 0.001	1.16 (1.09–1.24)	P = 0.001

Cox proportional hazards models were used to estimate HR and 95% CI

HR Hazards ratio, CI Confidence interval

Model 1 was unadjusted

Model 2 was adjusted for Age, Gender, Smoking, Alcohol, BMI

Model 3 was adjusted for Age, Gender, Smoking, Alcohol, BMI, Temperature, RR, Spo2, PLT, Lactate, cTnT, LDL, Glu, LDH, Potassium, Stroke, HTN, ACEI/ARB, Betablocker, Statin, NOAC

Moreover, RCS regression curves indicated that in the fully adjusted Model 3, some nutritional indicators had complex nonlinear dose-response relationships with 30-day ACM risk, rather than the simple linear associations suggested by Cox proportional hazards regression analysis (Fig. 2). Specifically, CONUT, GNRI, HALP, NPAR, and RAR had linear relationships with 30-day ACM risk (P for nonlinearity > 0.05); while ALI, BAR, CAR, LAR, NAR, and PNI had nonlinear relationships (P for nonlinearity < 0.05) (Fig. 2). MAR had no statistical significance (P for overall = 0.1795). In the 365-day ACM, MAR and NAR had no statistical significance (P for overall > 0.05), while all other nutritional indicators maintained consistent linear/nonlinear patterns with their respective 30-day mortality risks (Supplementary Fig. 4). This reflects the complexity and homogeneity of albumin-related nutritional indicators.

Fig. 2.

RCS regression analysis of 30-day ACM

Exploratory two‑piecewise regression suggested potential inflection points for ALI, BAR, CAR, LAR, NAR, PNI, and RAR in relation to 30‑day all‑cause mortality (P for likelihood ratio test < 0.05), with inflection points at 17.73, 7.29, 3.41, 0.54, 2.32, 42.29, and 4.08, respectively (Supplementary Table 7. For 365‑day all‑cause mortality, potential inflection points were also identified for ALI, BAR, CAR, LAR, PNI, and RAR (Supplementary Table 8). These estimates were derived from the present cohort and should not be interpreted as validated clinical cut‑off values.

Incremental value analysis

Incremental predictive value analyses showed that, compared with the reference model including albumin alone, most nutritional–inflammatory indices improved the prediction of 30-day all-cause mortality to varying degrees. LAR and BAR showed the greatest improvements, increasing the AUC from 0.5511 to 0.7191 and 0.6993, respectively (both P < 0.001) (Table 3). The IDI and NRI were 0.0957 and 0.5902 for LAR, and 0.0642 and 0.5047 for BAR, respectively (all P < 0.001), indicating significant improvements in model discrimination and risk reclassification. For 365-day all-cause mortality, BAR and RAR demonstrated the greatest incremental value, with AUCs increasing to 0.7173 and 0.6932, respectively (both P < 0.001) (Supplementary Table S9). The IDI and NRI were 0.0955 and 0.6015 for BAR, and 0.1036 and 0.5548 for RAR, respectively. Overall, LAR was more informative for short-term mortality prediction, BAR showed consistently strong performance for both short- and long-term outcomes, and RAR provided the greatest improvement in long-term risk discrimination.

Table 3.

Incremental predictive value of albumin-based nutritional-inflammatory indices over albumin alone for 30-day all-cause mortality

Model	AUC	ΔAUC (95% CI)	P-value	Continuous IDI (95% CI)	P-value	Continuous NRI (95% CI)	P-value
Albumin	0.5511	Ref.		Ref.		Ref.
ALI	0.5817	0.0305 (0.0141, 0.0470)	P < 0.001	0.0034 (0.0014, 0.0053)	P < 0.001	0.2334 (0.1459, 0.3183)	P < 0.001
BAR	0.6993	0.1482 (0.1103, 0.1740)	P < 0.001	0.0642 (0.0526, 0.0764)	P < 0.001	0.5047 (0.4114, 0.5946)	P < 0.001
CAR	0.6562	0.1050 (0.0797, 0.1304)	P < 0.001	0.0196 (0.0132, 0.0262)	P < 0.001	0.4154 (0.3257, 0.5063)	P < 0.001
CONUT	0.6634	0.1123 (0.0765, 0.1480)	P < 0.001	0.0631 (0.0521, 0.0741)	P < 0.001	0.4827 (0.3916, 0.5735)	P < 0.001
GNRI	0.6570	0.1059 (0.0697, 0.1420)	P < 0.001	0.0610 (0.0503, 0.0721)	P < 0.001	0.4344 (0.3412, 0.5276)	P < 0.001
HALP	0.6048	0.0537 (0.0202, 0.0872)	P = 0.002	0.0233 (0.0169, 0.0297)	P < 0.001	0.2852 (0.1920, 0.3798)	P < 0.001
LAR	0.7191	0.1680 (0.1355, 0.1983)	P < 0.001	0.0957 (0.0801, 0.1125)	P < 0.001	0.5902 (0.5016, 0.6822)	P < 0.001
MAR	0.5630	0.0119 (-0.0126, 0.0364)	P = 0.340	0.0067 (0.0028, 0.0113)	P = 0.002	0.1787 (0.0902, 0.2757)	P < 0.001
NAR	0.6233	0.0722 (0.0391, 0.1053)	P < 0.001	0.0353 (0.0261, 0.0450)	P < 0.001	0.3830 (0.2920, 0.4774)	P < 0.001
NPAR	0.6437	0.0926 (0.0579, 0.1273)	P < 0.001	0.0477 (0.0375, 0.0578)	P < 0.001	0.4428 (0.3519, 0.5325)	P < 0.001
PNI	0.6421	0.0910 (0.0578, 0.1242)	P < 0.001	0.0266 (0.0205, 0.0323)	P < 0.001	0.4341 (0.3396, 0.5238)	P < 0.001
RAR	0.6728	0.1217 (0.0872, 0.1561)	P < 0.001	0.0616 (0.0497, 0.0736)	P < 0.001	0.4750 (0.3865, 0.5659)	P < 0.001

Subgroup analysis

Subgroup analyses were performed for the two nutritional indicators most strongly associated with ACM. For 30-day mortality, the effect of LAR was consistent across all subgroups, including age, BMI, gender, smoking, CKD, HTN, and DM, with no significant interactions (Fig. 3). In contrast, BAR showed heterogeneity: while its association with 30-day mortality was consistent across age, smoking, HTN, and DM, stronger associations were observed in patients with BMI > 28 [1.19 (1.07–1.33), P = 0.002 vs. 1.05 (0.96–1.14), P = 0.316], women [1.28 (1.15–1.44), P < 0.001 vs. 1.05 (0.96–1.14), P = 0.287], and those without CKD [1.18 (1.07–1.30), P = 0.001 vs. 1.03 (0.93–1.14), P = 0.580], with significant interactions (P for interaction = 0.009, 0.004, and 0.020, respectively) (Fig. 3). For 365-day ACM, RAR showed consistent effects across all subgroups. In contrast, BAR demonstrated consistent associations across age, gender, smoking, HTN, and DM subgroups, but was significantly associated with 365-day mortality in patients with BMI > 28 [1.26 (1.15–1.38, P < 0.001) vs. 1.07 (1.00–1.15, P = 0.051)] and in those without CKD [1.22 (1.12–1.33, P < 0.001) vs. 1.07 (0.99–1.16, P = 0.097)], whereas no significant associations were observed in patients with BMI ≤ 28 or with CKD, with significant interaction effects (P for interaction = 0.001 and < 0.001, respectively) (Supplementary Fig. 5).

Fig. 3.

Subgroup analysis of 30-day all-cause mortality (A: LAR, B: BAR)

The associations between secondary clinical outcomes and the nutritional indicators

Univariable logistic regression showed that, among the 12 nutritional indices, only MAR was not significantly associated with NOAF (Supplementary Table 10). In multivariable logistic analysis, after full adjustment for confounders, CONUT, GNRI, HALP, NAR, and NPAR were significantly associated with NOAF in patients with AMI only in Group 4; ALI and PNI levels were negatively correlated with the incidence of NOAF, whereas BAR, CAR, LAR, and RAR were positively correlated (Supplementary Table 14). For NOVA, only BAR, CAR, CONUT, GNRI, NPAR, and RAR were significantly associated with NOVA in critically ill patients with AMI (Supplementary Table 12). In the fully adjusted model, RAR and CONUT were significant only in Group 4; BAR, CAR, and NPAR were positively associated with NOVA, while GNRI was negatively associated (Supplementary Table 15). No statistically significant associations were observed for AHF or mechanical complications. However, only 25 mechanical complication events occurred, and the corresponding regression results should therefore be interpreted as exploratory because of the limited statistical power and potential instability of the estimates (Supplementary Tables 11 and Supplementary Table 13).

Sensitivity analysis

In the sensitivity analyses, participants with missing LDL values were retained, and the missing data were handled using multiple imputation. Cox regression analyses for the primary outcomes were then repeated. The associations of the nutritional–inflammatory indices with the primary outcomes were generally consistent with those observed in the main analyses (Supplementary Tables S16 and S17). The findings remained robust after removing covariates that overlapped with the components of the corresponding indices (Supplementary Tables S18 and S19). Furthermore, additional adjustment for cardiogenic shock, percutaneous coronary intervention, vasopressor use, and mechanical ventilation did not materially alter the results (Supplementary Tables S18 and S19). Further adjustment for SOFA score in the fully adjusted model did not materially change the associations of nutritional‑inflammatory indices with 30‑day and 365‑day all‑cause mortality (Supplementary Tables S20 and S21).

Discussion

This study systematically discussed and compared the value of various albumin-based nutritional and inflammatory indices in predicting adverse outcomes in critically ill AMI patients for the first time. Higher levels of GNRI, HALP, and PNI were associated with a lower risk of ACM, while higher levels of BAR, CAR, CONUT, LAR, NPAR, and RAR were associated with increased mortality risk. Among them, CONUT, GNRI, HALP, NPAR, and RAR showed linear relationships with mortality risk, while ALI, BAR, CAR, LAR, NAR, and PNI exhibited nonlinear relationships. Exploratory nonlinear analyses suggested several data-derived inflection points. These findings may help generate hypotheses regarding nonlinear risk patterns but should not be interpreted as definitive clinical thresholds. Among the indices evaluated, LAR showed the highest discriminatory performance for 30-day all-cause mortality, whereas BAR showed the highest discriminatory performance for 365-day all-cause mortality. Additionally, many nutritional indices were associated with NOAF and NOVA during AMI hospitalization. However, no associations were observed between nutritional indices and AHF or mechanical complications.

Albumin-related nutritional indices have been widely used in various populations. Jia et al. used eight nutritional and inflammatory indices to identify significant interactions between elevated levels of these indices and ACM in Parkinson’s disease, emphasizing the importance of considering both inflammation and nutritional status when assessing mortality risk in Parkinson’s disease patients, with PNI and RAR being particularly important predictors of ACM [17]. Jiang et al. used five nutritional and inflammatory assessment tools to show that increased malnutrition risk was associated with poor clinical outcomes in elderly ischemic stroke patients, with PNI being the most effective in predicting poor prognosis one year later [18]. Xu et al. analyzed six nutritional and inflammatory indices related to the risk of chronic obstructive pulmonary disease and ACM, finding that different indices had varying impacts on outcomes, with ALI potentially providing more effective predictive value for risk and ACM [19].

In the cardiovascular field, the association between albumin-based nutritional indices and adverse outcomes has also attracted considerable attention, with generally consistent findings. In patients with acute myocardial infarction, several nutritional indices, including GNRI, HALP, NPAR, PNI, and CONUT, have been closely associated with both short- and long-term mortality [20–24]. In patients with heart failure, lower PNI or GNRI values are associated with higher in-hospital and 1-year mortality and longer hospital stays. Further analyses suggest that nutritional status may modify the association between BMI and prognosis; therefore, albumin-based nutritional indices may provide better risk stratification than BMI alone [25, 26]. However, no study has yet reached a unified conclusion on the best index and its effect. Building on previous research, this study summarized and analyzed both common and uncommon albumin-related nutritional indices. Specifically, for short-term mortality in AMI patients, LAR demonstrated the highest overall incremental predictive value, with an AUC of 0.719. This can be explained by the fact that AMI patients typically face intense acute inflammatory responses and severe stress-induced metabolic disorders in the short term, which are the main determinants of short-term prognosis [27]. Lactate, a product of anaerobic metabolism, usually increases during tissue hypoxia or ischemia, which is common in the acute phase of AMI. Elevated lactate levels can lead to myocardial injury, ventricular dysfunction, and ultimately poor clinical outcomes [28]. Additionally, low albumin levels indicate poor nutritional status, and acute inflammatory reactions can further reduce albumin levels [29]. LAR, which combines lactate and albumin, integrates inflammatory and nutritional factors and can more comprehensively reflect the body’s inflammatory/nutritional status, thereby more accurately predicting adverse outcomes. On the other hand, for long-term mortality, BAR demonstrated the highest overall incremental predictive value, with an AUC of 0.717. In the long-term survival of AMI patients, maintaining good nutritional status and stable hemodynamics is more important as the disease progresses to the recovery or chronic phase, both of which significantly improve subsequent quality of life and survival time [30]. Elevated BUN levels indicate deteriorating hemodynamics and renal perfusion injury [31], and studies have shown a significant association between BUN and long-term poor prognosis in AMI and AHF [32, 33]. Moreover, good nutritional status helps maintain metabolic homeostasis and reduce inflammation levels, avoiding persistent metabolic stress and acute inflammatory reactions [34]. Theoretically, compared to assessing BUN and albumin separately, BAR can more accurately predict adverse outcomes in AMI because it reflects liver and kidney function, protein metabolism, inflammation levels, nutritional status, and hemodynamic performance [35]. Additionally, we identified a nonlinear relationship between BAR and long-term ACM. Specifically, when BAR was less than 7.52, each unit increase was associated with a 24% increase in long-term mortality risk (HR: 1.242, 95% CI: 1.199–1.287, P < 0.001). However, when BAR was 7.52 or higher, further increases in BAR were associated with a smaller rise in long-term mortality risk (HR: 1.015, 95% CI: 1.007–1.022, P = 0.001), suggesting a possible saturation effect. This is similar to the findings reported by Wang et al. [35]. This pattern may indicate a possible plateauing association within the present cohort. However, the estimated inflection point of 7.52 was data-driven and should not be regarded as a clinical cut-off until it has been externally validated. Lactate, BUN, and albumin are all simple items in routine tests, easily accessible and relatively objective. These indices may serve as adjunctive assessment tools to help clinicians better understand the current inflammatory and nutritional status of patients and improve the identification of high-risk individuals, thereby providing a reference for individualized risk management and improving patient outcomes. Nevertheless, the AUC values of LAR and BAR indicated only moderate discrimination. Therefore, these indices should not be interpreted as stand-alone clinical decision tools but may serve as readily available adjuncts to established clinical risk-assessment approaches.

In terms of new-onset arrhythmias, a meta-analysis indicated that low serum albumin levels are associated with an increased risk of atrial fibrillation [36]. A large cohort study involving 320,111 cases of ischemic stroke showed that low albumin levels are associated with early NOVA [37]. In this study, albumin-related nutritional indices were associated with NOAF and NOVA during AMI hospitalization, consistent with previous studies. However, no association was found between albumin-based nutritional indices and mechanical complications after AMI. This is contrary to Zhang et al.‘s study and may be related to the types of mechanical complications included in this study, which included mitral valve prolapse or rupture, ventricular septal defect, and cardiac rupture, while Zhang et al. included ventricular aneurysm [38]. However, given the small number of events, this finding may reflect limited statistical power and should not be interpreted as evidence of no association. In terms of in-hospital AHF, no significant association was observed between nutritional indices and AHF in this study, suggesting that nutritional-inflammation imbalance may primarily mediate adverse events through electrophysiological remodeling and multi-organ failure rather than structural rupture. Notably, this study found that nutritional indices with threshold effects had similar inflection points in their associations with clinical endpoints. A similar outcome was observed in a cohort study of ventilator-associated pneumonia, where PNI had similar inflection points in short-term and long-term ACM associations, suggesting that there may be a common underlying pathophysiological mechanism. However, because the strict temporal sequence between the first laboratory measurements and the true onset time of secondary outcomes during hospitalization could not be reliably determined, reverse causation cannot be completely excluded. Therefore, the logistic regression results for secondary outcomes should be interpreted as exploratory associations rather than prospective risk estimates or causal effects.

The mechanisms underlying the associations of albumin-based nutritional–inflammatory indices with all-cause mortality and new-onset arrhythmias in patients with AMI remain uncertain. Albumin has physiological antioxidant properties and may reduce oxidative stress through its free thiol groups [39]. In patients with AMI, reduced albumin levels may reflect inadequate nutritional intake, enhanced protein catabolism, systemic inflammation, impaired hepatic synthesis, or increased vascular permeability [40]. These abnormalities often coexist with hemodynamic instability and organ dysfunction and may therefore partly explain the observed association between low albumin-based indices and mortality. Oxidative stress has also been associated with platelet activation, impaired fibrinolysis, endothelial dysfunction, myocardial injury, and electrical instability; however, these pathways were not directly evaluated in the present study [37, 41]. Similarly, systemic inflammatory mediators, including tumor necrosis factor-α and interleukin-6, may contribute to endothelial activation, leukocyte recruitment, metabolic disturbance, and progressive organ dysfunction [42]. Rather than representing a single biological pathway, the evaluated indices may integrate several dimensions of illness severity, including nutritional depletion, inflammatory burden, renal dysfunction, and tissue hypoperfusion. Accordingly, they should primarily be interpreted as composite prognostic markers. Their potential mechanistic or causal roles require confirmation in studies incorporating direct measurements of inflammation, oxidative stress, and organ injury.

Despite these findings, several limitations should be acknowledged. First, the retrospective design precludes causal inference. Second, although multiple known confounders were adjusted for, residual confounding from unmeasured lifestyle factors, such as dietary habits, may remain. AMI was identified using ICD‑9 and ICD‑10 codes, and some cases were coded as unspecified AMI, which prevented reliable distinction between STEMI and NSTEMI; time to reperfusion was also unavailable. Killip classification was not reliably available in MIMIC‑IV. Although we further adjusted for the SOFA score, residual confounding related to AMI‑specific severity cannot be completely excluded. In addition, arrhythmic events were identified based on routine bedside documentation without independent electrocardiographic adjudication. Standardized nutritional assessment tools, including the Mini Nutritional Assessment–Short Form (MNA‑SF) and Nutritional Risk Screening (NRS), were not available in the database; therefore, prospective studies incorporating such assessments are needed. Although multiple imputation was used to address missing data, potential bias related to the missing‑data mechanism cannot be completely eliminated. Furthermore, only baseline albumin levels were evaluated, without accounting for dynamic changes during hospitalization or subsequent admissions. Future prospective studies combining standardized nutritional assessment with serial albumin measurements are warranted to clarify the association between changes in nutritional status and clinical outcomes. Finally, this study was based on critically ill AMI patients from a single U.S. tertiary academic center, so caution is needed when generalizing the findings to non-critically ill populations, other healthcare systems, or populations with different racial and geographic backgrounds. External validation in large-scale, multicenter, and diverse cohorts is warranted in the future.

Conclusion

The study results indicate that nutritional–inflammatory imbalance was associated with adverse outcomes in critically ill patients with AMI. Both in the short and long term, low ALI, GNRI, HALP, and PNI, as well as high BAR, CAR, CONUT, LAR, NPAR, and RAR, consistently suggest a stepwise increase in mortality risk. Among the indices evaluated, LAR and BAR showed the greatest discriminatory performance for short- and long-term all-cause mortality, respectively. These readily available indices may complement conventional clinical risk assessment but should not be used as stand-alone clinical decision tools. The associations between several nutritional–inflammatory indices and mortality were nonlinear, and exploratory analyses identified several data-derived inflection points. These estimates require external validation and should not be interpreted as established clinical thresholds.

Authors’ contributions

HW, XW and YH wrote the text of the main manuscript, carried out the data data organization and analysis and the production of tables and images. HH: revised the manuscript. All authors reviewed the manuscript.

Funding

The author(s) declare that no financial support was received for the research, authorship, and/or publication of this article.

Data availability

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding authors.

Declarations

Ethics approval and consent to participate

Ethical approval was not required for the studies involving humans because this study utilized data from the MIMIC-IV database, which was approved for research use by the Institutional Review Boards (IRBs) of the Massachusetts Institute of Technology (MIT) and Beth Israel Deaconess Medical Center (BIDMC). The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation was not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and institutional requirements because only de-identified, publicly available patient data were used.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.