Over recent decades, cardiovascular diseases represent a primary cause of mortality worldwide. In the year 2021, these conditions were responsible for an estimated 20.5 million deaths worldwide. Of these, around 8.99 million were due to ischemic cardiovascular conditions such as acute myocardial infarction (AMI) [1, 2]. AMI is a particularly severe and frequent presentation of ischemic heart disease. The incidence of AMI increases markedly with age, affecting as many as 9.5% of individuals over the age of 60 years [3]. Critically ill patients in the intensive care unit (ICU) often exhibit a range of intricate health issues and coexisting risk factors. Studies indicate that approximately 4%–14% of ICU patients experience AMI during hospitalization [4]. Despite these observations, there is still only limited research on prognostic indicators and risk stratification in critically ill patients with AMI. It is therefore imperative to conduct additional studies allowing a deeper understanding of this high-risk cohort. Timely recognition and proper management of identified risk factors are essential for lowering the mortality rate in this patient population.

AMI involves complex immunological and inflammatory responses. Previous studies have suggested that combined biomarkers, including the Systemic Immune-Inflammation Index [5], Systemic Inflammatory Response Index [6], neutrophil-to-lymphocyte ratio (NLR) [7], Prognostic Nutritional Index [8], and Controlling Nutritional Status score [9], may have superior prognostic value for AMI compared with single inflammatory or nutritional markers alone [10].

The Hemoglobin, Albumin, Lymphocyte, and Platelet (HALP) score was first proposed as a prognostic tool for various types of cancers [11, 12, 13]. Recent studies have also demonstrated prognostic utility for HALP in various cardiovascular conditions, such as acute heart failure [14], coronary artery disease [15], patients undergoing percutaneous coronary intervention (PCI) [16], and individuals recovering from coronary artery bypass grafting (CABG) [17]. This evidence has increased the acceptance of HALP as a composite marker reflecting both systemic inflammation and nutritional state in the field of cardiovascular medicine. However, only limited research has been directed at specifically evaluating the prognostic relevance of the HALP score in individuals diagnosed with AMI. This gap is particularly pronounced for the high-risk subpopulation of critically ill patients with AMI, whose complex pathophysiology and management in ICU necessitates more precise risk stratification tools. Considering that AMI is accompanied by significant immune-inflammatory activation and nutritional disturbances [18, 19], we hypothesized that the HALP score may be a robust prognostic tool for predicting outcomes in this cohort. Consequently, the aim of our study was to evaluate the association between HALP score and all-cause mortality among critically ill patients with AMI.

In this study, we conducted a comprehensive examination of the link between HALP score and all-cause mortality in critically ill AMI patients. We found a significant association between higher HALP score and reduced risk of mortality at both 90 days and 365 days follow-up periods. Multivariable Cox regression models revealed that individuals with HALP scores in the upper quartile (Q4) had a significantly lower risk of death compared to those in Q1, with an HR of 0.68 for 90 days mortality and 0.66 for 365 days mortality (both p 0.05). Threshold effect analyses were performed using both standard and two-piece Cox models, with an inflection point identified at a HALP score of 19.41 for both time points. Below this identified threshold, each one-point rise in HALP score corresponded to a 2.4%–2.7% decrease in mortality risk. However, no significant relationship was observed when the HALP score exceeded the cutoff. Additionally, machine learning models incorporating the HALP score and other clinical variables demonstrated strong predictive performance for 90 days mortality, with the highest AUC reaching 0.78. Furthermore, our additional analyses demonstrated that a model incorporating the HALP score provides significant incremental value in risk prediction and clinical utility over established scores like the CCI and GCS, as evidenced by NRI and DCA results. These results highlight the value of the HALP score as a predictive marker and its applicability in developing targeted interventions to reduce mortality in critically ill AMI patients.

Immune-inflammatory mechanisms are pivotal in driving the progression of AMI and determining its prognosis. Following the onset of AMI, a robust immune-inflammatory response is triggered, resulting in substantial release of damage-associated molecular patterns. These facilitate the recruitment and infiltration of neutrophils, monocytes, and macrophages into the infarcted myocardium [22, 23]. During this phase, neutrophils exacerbate the local myocardial injury by releasing proteolytic enzymes and reactive oxygen species via degranulation, thereby expanding the infarct size and initiating maladaptive left ventricular remodeling [24]. Macrophages further amplify local inflammation, thereby exacerbating ventricular dilatation and dysfunction [25]. Concurrently, hospitalized AMI patients frequently exhibit nutritional deficits, such as hypoalbuminemia and anemia, which impair immune function, reduce resistance to inflammatory damage, and diminish the capacity for tissue regeneration [26]. Studies have demonstrated that malnutrition significantly increases mortality in AMI patients and is an independent predictor of poor outcomes [27]. Thus, the exaggerated immune-inflammatory response following AMI directly exacerbates myocardial injury and adverse remodeling, while concurrent nutritional deficiencies weaken the body’s immune defence and repair mechanisms. Together, these synergistic effects contribute substantially to poor patient outcomes.

The HALP score functions as a composite biomarker that simultaneously captures both immune-inflammatory activity and nutritional condition. Our findings demonstrate a nonlinear relationship of the HALP score with mortality in critically ill AMI patients. This was identified as an inverse “L-shaped” curve by RCS analysis. A significant inflection point was detected at a HALP score of 19.41, below which the risk of mortality declined sharply with each unit increase in HALP. After this threshold, the risk of mortality stabilized. The threshold effect provides a clinically actionable cutoff, allowing stratification of ICU patients into low- and high-risk groups at the time of admission. Such early risk stratification can facilitate timely nutritional support, anti-inflammatory interventions, and intensive monitoring. The observed nonlinear relationship underscores the combined impact of immune-inflammatory responses and nutritional status on patient prognosis. Firstly, a lower HALP score suggests dual impairment of immune function and nutritional status. Lymphopenia diminishes the body’s capacity to effectively regulate and suppress inflammation. Previous studies have indicated that AMI patients with lower lymphocyte counts and higher NLR have a significantly increased risk of long-term mortality. Furthermore, impaired peripheral T-lymphocyte function has been shown to exacerbate myocardial ischemia-reperfusion injury [28, 29]. Concurrently, platelets not only contribute to coronary artery and microvascular thrombosis during AMI, but also exacerbate inflammation and reperfusion injury. This inflammatory-thrombotic interaction is recognized as a critical contributor to adverse outcomes in AMI [30, 31]. Increased platelet levels typically indicate a cytokine-driven acute-phase response that promotes inflammation, most notably interleukin-6, leading to a hypercoagulable and pro-inflammatory state [32]. Compared to patients with moderate platelet counts (250–349 K/µL), those with higher counts ($\ge$350 K/µL) were reported to show increased overall mortality following AMI [33]. Secondly, hypoalbuminemia, as included in the HALP score, not only indicates malnutrition but also serves as a marker of increased inflammation and oxidative stress [34]. Clinical evidence consistently demonstrates that hypoalbuminemia independently predicts adverse outcomes following AMI [35]. To illustrate this, a cohort study of 7192 patients with acute coronary syndrome (ACS) found that individuals with a serum albumin level 3.5 g/dL at admission had significantly higher rates of both in-hospital death and heart failure [36]. Moreover, a lower HALP score typically coexists with reduced hemoglobin levels (anemia), directly impairing oxygen transport and tissue perfusion. Anemic states intensify myocardial hypoxia, potentially enlarging the infarct size and precipitating cardiac dysfunction. Severe anemia (hemoglobin 9 g/dL) has been shown to significantly increase short-term mortality among AMI patients, and a hemoglobin level of 9 g/dL is associated with an approximately 50% increase in the 120 days mortality risk [37]. At a mechanistic level, the robust prognostic power of the HALP score in critically ill AMI patients likely stems from the synergistic interplay among its components, which collectively reinforce a vicious cycle of inflammation, hypoxia, and impaired healing. For example, anemia-induced hypoxia can exacerbate inflammatory signaling via HIF-1$\alpha$ pathways [38, 39], while inflammation in turn suppresses erythropoiesis, further worsening anemia [40]. Concurrently, hypoalbuminemia not only signals depleted nutritional reserves, but also weakens antioxidant defenses and buffering capacity against inflammatory cytokine storms [41]. This effect is particularly pronounced in the setting of lymphocytopenia and subsequent immune dysregulation. The resulting systemic vulnerability predisposes patients to severe microvascular damage during ischemia–reperfusion and maladaptive cardiac remodeling, ultimately leading to adverse clinical outcomes.

Of note, our analysis of subgroups identified a significant interaction between the HALP score and age in relation to 365 days mortality. The protective effect of a high HALP score was markedly stronger in patients aged $\ge$70 years, but was not significant in the younger cohort. A likely explanation for this is that elderly patients, because of their diminished physiological reserves, are more vulnerable to the nutritional and immune insults captured by the HALP score. In contrast, mortality in younger patients may be driven by more aggressive pathophysiological factors that overshadow the HALP parameters. This finding highlights the utility of the HALP score as a particularly crucial long-term prognostic marker for risk-stratification of elderly, critically ill AMI patients.

In summary, a low HALP score reflects an impaired immune defense and poor nutritional reserves, resulting in uncontrolled inflammatory responses and oxidative stress. This weakened immune state further increases the patients’ susceptibility to complications such as infections, while hindering myocardial repair, delaying functional recovery, and ultimately worsening patient prognosis. Conversely, when the HALP score exceeds a certain threshold, it indicates a relatively favorable immune and nutritional status. Beyond this point, further increases in HALP score provide diminishing marginal benefit in terms of prognosis, representing a plateau in its predictive utility. Moreover, recent findings indicate that certain glucose-lowering drugs, such as GLP-1 receptor agonists and SGLT2 inhibitors, may provide cardioprotective effects in AMI, irrespective of their glycemic control function. These agents exhibit anti-inflammatory and endothelial-stabilizing properties, which may interact with nutritional and inflammatory pathways reflected in the HALP score [42]. Although not addressed in the present study, this evolving therapeutic landscape warrants further investigation. Compared to previous studies, our work offers several novel insights. Pannu [43] emphasized the theoretical advantages of HALP and CALLY as systemic indices to supplement traditional ACS risk models, but provided no primary data from critically ill cohorts. Yılmaz et al. [44] focused on elderly AMI patients ($\ge$75 years) undergoing PCI and identified HALP as a long-term predictor of mortality in a relatively small, elective cohort. In contrast, our study targeted critically ill AMI patients in the ICU setting and provided robust statistical modeling, including nonlinear and machine learning analyses. We identified a clinically actionable threshold (HALP = 19.41) to facilitate early risk stratification and personalized care.

Nonetheless, this study has several limitations that should be taken into consideration. First, a key limitation is the selection bias associated with the exclusion of 73.8% of patients who lacked HALP data. Because the patients included in the study cohort were more unwell and had higher mortality, our findings on the prognostic value of the HALP score apply mainly to this high-risk group and may not generalize to less severe AMI populations. Second, due to a substantial amount of missing data for inflammatory biomarkers and lipid profiles in the database, these variables could not be incorporated into our analysis. This absence may limit a more comprehensive understanding of the immune and metabolic pathways involved in AMI, potentially affecting the completeness of our predictive models. Third, the retrospective and single-center design of the study is a further limitation, with the findings being susceptible to selection bias and residual confounding, thus restricting our ability to establish causality. Fourth, this was a retrospective, single-center study based on a U.S. tertiary academic hospital cohort, with patients limited to those admitted to the ICU for AMI. Caution is warranted when extrapolating the findings to AMI populations outside the ICU, in resource-limited settings, or in different healthcare systems. Furthermore, the findings may not fully apply to patients with specific clinical subtypes, such as ST-elevation or non-ST-elevation MI. Lastly, the HALP score in this study was calculated from lab values obtained during the initial 24 h of the patient’s stay in ICU. This may reflect acute physiological stress rather than chronic nutritional or inflammatory status. Therefore, caution is warranted when interpreting HALP as a modifiable biomarker. To overcome these shortcomings, future investigations should focus on large-scale, prospective studies conducted at multiple centers. Such a design would be instrumental in validating our findings across more heterogeneous AMI populations, including non-ICU patients and those from different healthcare systems. This would enhance the generalizability of our findings and reduce selection bias. Furthermore, additional studies should aim for the systematic collection of serial HALP measurements alongside a comprehensive panel of inflammatory and metabolic biomarkers. Such an approach is crucial for mitigating confounding factors and better elucidating the temporal dynamics and causal role of the HALP score in AMI prognosis.

The present study validates the HALP score as an independent predictor of mortality for patients with AMI. Moreover, machine learning models incorporating the HALP score showed strong performance in predicting mortality risk, further highlighting its potential utility in clinical decision-making. These results support use of the HALP score as a practical, economical, and objective tool for early risk stratification and outcome prediction in critically ill patients with AMI.

HALP, hemoglobin, albumin, lymphocyte, and platelet; AMI, acute myocardial infarction; ICU, intensive care unit; MIMIC, Medical Information Mart for Intensive Care; AUC, area under the receiver operating characteristic curve; RCS, restricted cubic spline; HR, hazard ratio; CI, confidence interval; RR, respiratory rate; SBP, systolic blood pressure; DBP, diastolic blood pressure; SPO₂, peripheral capillary oxygen saturation; Hb, hemoglobin; INR, international normalized ratio; PH, potential of hydrogen; PTT, partial thromboplastin time; WBC, white blood cell count; PCO₂, partial pressure of carbon dioxide; Cr, creatinine; BUN, blood urea nitrogen; GLU, glucose; PO₂, partial pressure of oxygen; AF, atrial fibrillation; CA, cancer; CKD, chronic kidney disease; CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; MV, mechanical ventilation; APS-III, Acute Physiology Score III; CCI, Charlson Comorbidity Index; GCS, Glasgow Coma Scale; SAPS-II, Simplified Acute Physiology Score II; SOFA, Sequential Organ Failure Assessment; SII, Systemic Immune-Inflammation Index; SIRI, Systemic Inflammatory Response Index; NLR, neutrophil-to-lymphocyte ratio; PNI, Prognostic Nutritional Index; CONUT, Controlling Nutritional Status; DAMPs, damage-associated molecular patterns; ROS, reactive oxygen species; IL-6, interleukin-6; LASSO, least absolute shrinkage and selection operator; ENet, elastic net; LightGBM, Light Gradient Boosting Machine; MLP, multilayer perceptron; RF, random forest; XGBoost, extreme gradient boosting; ROC, receiver operating characteristic; DCA, decision curve analysis; VIF, variance inflation factor.

This study analyzed publicly available datasets from the MIMIC-IV v3.1 database (http://physionet.org/content/mimiciv/3.1/).

Conceptualization: ZTC and YSW. Methodology: ZTC. Software: ZTC. Validation: HJJ, JL, and KLZ. Formal Analysis: ZTC and NJC. Data Curation: ZTC. Funding Acquisition: JC. Writing – Original Draft: ZTC and NJC. Writing – Review & Editing: HJJ, JL, KLZ, JTC, YSW, and JC. Visualization: JTC and JC. Supervision: YSW. All authors contributed to the conception and editorial changes in the manuscript. All authors have read and approved the final manuscript and agree to be accountable for all aspects of the work.

The study was carried out in accordance with the guidelines of the Declaration of Helsinki. The studies involving human participants were approved by the Institutional Review Boards (IRBs) of Beth Israel Deaconess Medical Center (Protocol No. 2001-P-001699/14) and the Massachusetts Institute of Technology (Protocol No. 0403000206). All studies were conducted in compliance with local laws and institutional guidelines. The ethics committee/IRB waived the requirement for written informed consent from participants or their legal guardians/next of kin, as the database used in this study anonymizes patient information, thus eliminating the need for informed consent.

This research was funded by Xiamen Municipal Bureau of Science and Technology, Grant No. 3502Z20224ZD1172.

The authors declare no conflict of interest.

Supplementary material associated with this article can be found, in the online version, at https://doi.org/10.31083/RCM43942.