Introduction

Acute Myocardial Infarction (AMI), the most catastrophic manifestation of coronary artery disease, typically results from acute plaque rupture or sustained coronary occlusion.¹ Despite advances in reperfusion therapies^2,3 such as the establishment of chest pain centers in China, which have significantly streamlined AMI care pathways,⁴ emerging data from the 2023 China Cardiovascular Health and Disease Report indicate a growing epidemiological burden, with national incidence reaching 79.7 per 100,000 person-years and mortality rates maintaining an ascending trend.⁵ Globally, AMI continues to dominate as the principal cause of mortality and disability-adjusted life years worldwide.⁶

Cerebral infarction (CI), the second-leading global cause of mortality and third-leading contributor to disability,⁷ constitutes a devastating yet underappreciated complication in AMI management. The novel clinical entity of cardio-cerebral infarction (CCI), defined by the co-occurrence of AMI and CI within 48 hours,⁸ presents unique therapeutic challenges. Although its incidence remains modest (3.2–4.7% of AMI admissions),⁹ meta-analyses reveal catastrophic outcomes: 2.3-fold elevated in-hospital mortality (95% CI 1.8–3.0) and 4.1-fold increased risk of major adverse cardiovascular events compared to isolated AMI.^10,11 Furthermore, survivors face substantially reduced functional capacity and quality of life.^12–14 These findings highlight the urgent need for early identification of AMI patients at high CI risk to implement preventive strategies.

Current research has identified several CI-associated risk factors in AMI populations, including cardiac dysfunction and systemic inflammation.^15,16 However, existing risk stratification paradigms remain fragmented, relying predominantly on isolated biomarkers rather than integrative prognostic modeling. Notably, no validated predictive tools currently enable individualized CI risk stratification in AMI patients. To address this gap, we aimed to construct a predict nomogram to discern who will experience in-hospital first-onset CI to help clinical decision-making.

Methods

Study Population

Data from patients diagnosed with AMI at the Cardiovascular Center of Meizhou People’s Hospital (Meizhou, China) between January and December 2022 were retrospectively collected. The diagnosis was confirmed according to the Fourth Universal Definition of Myocardial Infarction.¹⁷

The inclusion criteria were:(1) primary diagnosis of AMI; (2) complete baseline clinical documentation. The exclusion criteria were (1) pre-existing cerebrovascular diseases, such as prior cerebral infarction or intracranial hemorrhage; (2) active malignancy or terminal illness; (3) missing critical admission data (>20% of variables incomplete). This study was performed in accordance with the ethical standards of the Declaration of Helsinki and approved by the Human Ethics Committee of Meizhou People’s Hospital.

Data Collection

Demographic and Clinical Variables

We collected potential predictive factors from all patients, including age, sex, BMI, underlying diseases, clinical presentation, therapeutic intervention, left ventricular ejection fraction by echocardiography, and AMI subtype classification (ST-segment/non-ST-segment Elevation Myocardial Infarction). Standardized laboratory tests performed within 24 hours of admission, which included hematology, biochemistry, cardiac biomarkers, and inflammatory markers were collected. Derived indices, calculated as follows: Triglyceride-Glucose (TyG) index = ln [fasting triglycerides (mg/dL) × glucose (mg/dL)/2]; Neutrophil-to-lymphocyte ratio (NLR) = absolute neutrophils/absolute lymphocytes; Platelet-to-lymphocyte ratio (PLR) = platelet count/absolute lymphocytes, were collected.

Outcome Definition

The World Health Organization (WHO) has established a standardized coding and classification framework for CI in the International Classification of Diseases, 10th Revision (ICD-10).¹⁸ According to the WHO definition, the diagnosis of CI requires the integration of clinical manifestations, imaging examinations, and other auxiliary tests, while excluding other potential etiologies. In this study, we applied the WHO diagnostic principles for CI and the ICD-10 coding system to confirm cases of CI.

Statistical Analysis

The normal distribution of the continuous variables was assessed using the Kolmogorov–Smirnov test. Variables that followed a normal distribution were expressed as mean ± standard deviation (SD) and compared using the independent samples t-test. Variables that did not follow a normal distribution were expressed as median (interquartile range) and compared using the Mann–Whitney U-test. Categorical variables were expressed as frequency (percentage). Between-group comparisons for categorical data were performed using the chi-square (χ2) test.

The Least Absolute Shrinkage and Selection Operator (LASSO) regression analysis with 10-fold cross-validation was employed to reduce data dimensionality and identify the most valuable predictors (MVPs).¹⁹ The nomogram was developed by integrating all MVPs based on their regression coefficients in the binary logistic regression model. Subsequently, the calibration of the nomogram was evaluated using the calibration curve, accompanied by the Hosmer–Lemeshow test. The discrimination ability of the nomogram was assessed using the area under the receiver operating characteristic curve (AUC) and concordance index (C-index). Decision curve analysis (DCA) was performed to assess the net benefit of the nomogram for clinical decisions-making.²⁰ The optimal risk cutoff was determined by maximizing Youden’s index (sensitivity + specificity − 1).

A two-tailed P-value less than 0.05 was considered significant. The Statistical Package for R 4.3.3 software (R Foundation for Statistical Computing) was used for the complete statistical analyses.

Results

Baseline Characteristics

The study cohort consisted of 1,350 AMI patients, including 60 cases (4.44%) who developed first-onset CI during hospitalization (CI group) and 1,290 controls (non-CI group). The overall population showed a male predominance (76%, n = 1,031) with a median age of 68 years (interquartile range [IQR] 59–76). The demographic and clinical characteristics of the two cohorts are summarized in Table 1.

Table 1 General characteristics

Identification of Optimal Predictive Factors for First-Onset CI

Using LASSO regression with 10-fold cross-validation, the optimal hyperparameter λ was identified based on the minimum binomial deviance. Based on the optimal λ, five non-zero coefficients of preoperative features were selected as MVPs, including Killip classification, PCI therapy, C-reactive protein (CRP), blood urea nitrogen (BUN), and NLR (Figure 1).

Figure 1 Feature selection using the least absolute shrinkage and selection operator (LASSO) analysis with 10-fold cross-validation. Lambda (tuning parameter) selection of deviance in the LASSO regression based on the one standard error criteria (right dotted line) and the minimum criteria (left dotted line) (A). LASSO coefficient profiles of the candidate features. The intersecting curves represent the number of features retained at that log (lambda) value, and six predictors with nonzero coefficients were selected according to the one standard error criteria (B).

The odds ratios (95% CI) of the five MVPs were 1.4238(1.0513–1.9319), 0.2894(0.1638–0.5099), 1.0063(1.0021–1.0103), 1.0317(0.9941–1.0671), and 1.0219 (0.99–1.0525), as shown in Table 2. The list of candidate variables and their standardized coefficients were documented in Table 3.

Table 2 Details of Logistic Regression Analysis based on LASSO Screening Variables

Table 3 Variables Selected by LASSO Regression with Coefficient

Development of the Nomogram

Based on the coefficients of the five predictors, the nomogram was constructed using the “rms” package in R software (Figure 2).

Figure 2 Nomogram for predicting the risk of first-onset CI in patients with AMI.

To predict the risk of in-hospital first-onset CI in AMI patients using the nomogram, the total score is calculated by summing the individual scores of the five predictors. The risk of CI is then determined by projecting a vertical line downward from the total score.

Multiple Validations of the Nomogram

In the internal validation, the nomogram demonstrated a high accuracy in predicting CI in AMI patients, with an AUC of 0.804 (95% CI: 0.749–0.859; Figure 3A). Using the bootstrap internal resampling method, the C-index was 0.7917 (95% CI: 0.7134–0.870). The C-index obtained through ten-fold cross-validation was 0.7959. The Hosmer–Lemeshow test result (χ² = 4.6377, P = 0.7955) and a close-to-ideal calibration curve indicated good calibration of the nomogram (Figure 3B). DCA demonstrated clinically meaningful net benefits across a threshold probability range of 10% to 80%, suggesting robust utility of the nomogram for risk-stratification in real-world practice (Figure 3C).

Figure 3 Evaluation of the nomogram. The ROC curve (A), the calibration plot (B), and the decision curve analysis (C). In calibration plots, the calibration curve is expected to fall along the ideal line corresponding to a perfectly calibrated nomogram; the red line represents the apparent accuracy of the nomogram without correction for overfitting, while the blue line represents the bootstrap-corrected nomogram. In DCA plots, “all” refers to the assumption that all patients experienced first-onset CI, while “none” assumes that no patient experienced first-onset CI.

Determination of the Optimal Cutoff Value for the Nomogram

Based on data from all 1,350 patients enrolled in this study, as per the rule of the most balance between sensitivity and specificity, the optimal cutoff value of the nomogram was identified as 0.035 (sensitivity 0.783 and specificity 0.711). In other words, under the conditions of this study, patients with a nomogram probability of ≥0.035 were classified as high-risk patients with CI and should receive focused attention and targeted interventions, such as timely PCI treatment, to effectively prevent disease progression and improve patient outcomes.

Discussion

This study developed the first predictive tool to identify patients with AMI at risk of CI. The primary advantage of our model is its ability to enable physicians in resource-limited settings to predict the risk of in-hospital first-onset CI in AMI patients using physical examinations and readily available laboratory indices. Based on the nomogram results, we established an optimal cutoff value to assist clinicians in determining additional treatment strategies.

The co-occurrence of AMI and CI represents a critical clinical scenario with profound implications for patient outcomes. The reported incidence rates of this complication vary widely across studies, ranging from 0.4% to 12.7%. This variation may be attributed to differences in study populations, diagnostic criteria, and methodological approaches. Notably, Chin et al reported the highest incidence at 12.7% in their comprehensive analysis,²¹ while the GRACE trial reported a more conservative figure of 0.9% for in-hospital stroke among acute coronary syndrome patients.¹² Intermediate values were reported by Kajermo et al (2.1% within 30 days post-AMI)¹⁶ and Hachet et al (1.25% during hospitalization).²² In our study of 1,350 AMI patients, we identified 60 cases of concurrent CI during hospitalization, resulting in an incidence rate of 4.44%. This higher incidence may be attributed to our stringent inclusion criteria, which excluded patients with prior CI, cerebral hemorrhage, or malignant tumors to minimize confounding factors and improve study rigor. As a result, our study population was more homogeneous, which may have contributed to the higher observed incidence rate.

Previous studies have identified various risk factors for CI in AMI patients, with a predominant focus on non-laboratory indicators. Hachet et al identified female sex, history of transient ischemic attack or stroke, new-onset atrial fibrillation, left ventricular ejection fraction, and CRP as independent predictors.²² Mohammed et al identified hypertension as the most prevalent risk factor, followed by diabetes and atrial fibrillation.¹¹ Hurskainen et al emphasized the significance of paroxysmal atrial fibrillation/flutter, stroke history, left main coronary artery occlusion, coronary artery disease severity, and high Killip classification.²³ Our findings are consistent with these studies, showing significant differences in diabetes mellitus, atrial fibrillation, cardiogenic shock, acute heart failure, Killip classification, and reduced ejection fraction between patients with and without concurrent CI.

In addition to these established factors, our study identified several laboratory indices with statistically significant differences, including B-type natriuretic peptide (BNP), cardiac enzymes, uric acid, creatinine, BUN, total protein, albumin, prealbumin, fasting glucose, complete blood count parameters, TyG index, NLR, and PLR. Although these indices have been individually linked to AMI or CI, their combined analysis in our study provides a more comprehensive risk assessment. The inclusion of these laboratory parameters represents a significant advancement in risk stratification, as they offer objective and measurable indicators that are easily accessible in clinical practice.

Using LASSO regression analysis, we identified five optimal predictors for AMI-associated CI: Killip classification, PCI treatment, CRP, BUN, and NLR. These predictors were integrated into a nomogram, which demonstrated excellent predictive performance.

The association between higher Killip classification (III–IV) and increased stroke risk has been well-documented,^23,24 likely due to severe myocardial damage causing systemic circulation disturbances and cerebral hypoperfusion. Our findings that PCI treatment as a protective factor highlights the importance of timely coronary reperfusion in stroke prevention. Early revascularization has been demonstrated to reduce stroke risk in AMI patients,²⁵ potentially by improving myocardial function, reducing arrhythmias, and stabilizing of cerebral hemodynamics.The inclusion of CRP in our model is supported by its established role as a biomarker for stroke risk^26,27 and its identification as an independent predictor of in-hospital CI in AMI patients.²²

Elevated BUN levels have been consistently linked to an increased risk of stroke and ischemic stroke.²⁸ Moreover, elevated BUN levels have been associated with to correlate with poor prognosis in CI.²⁹ The underlying mechanism may involve the disruption of glucose homeostasis,³⁰ which can subsequently lead to the development of diabetes,³¹ thereby increasing the risk of CI. In addition, BUN levels are influenced by various factors, including protein intake, corticosteroid use, gastrointestinal bleeding, and dehydration,³² all of which reflect the body’s metabolic status. Notably, conditions such as dehydration or impaired renal function are also associated with an increased risk of stroke.³³ In our study, BUN levels were significantly higher in patients with CI compared to those without, consistent with previous findings. This consistency supports the validity of including BUN as a predictive factor in our model.

Emerging clinical studies have consistently identified elevated NLR levels in acute ischemic stroke patients compared to healthy controls.^34–37 This hematological phenomenon is mechanistically rooted in the robust inflammatory cascade triggered by CI.³⁸ Within hours post-infarction, ischemic brain tissue releases chemokines that recruit peripheral neutrophils to the lesion site, followed by lymphocyte infiltration.³⁹ Neutrophils dominate early pro-inflammatory responses through reactive oxygen species production and protease release,⁴⁰ while lymphocytes exert delayed immunoregulatory effects.⁴¹ The clinical elevation of NLR precisely mirrors the temporal dynamics of post-infarction inflammation, characterized by an acute neutrophilic surge peaking at 1–3 days⁴² followed by progressive lymphopenia stemming from glucocorticoid-induced apoptotic depletion.⁴³ Beyond its predictive value for infarction risk^44,45 and clinical prognosis,^37,46,47 NLR quantifies neuroinflammation-mediated secondary injury intensity, demonstrating superior sensitivity to traditional biomarkers in CT-negative minor stroke cases.⁴⁸ This ratio thus provides a clinically actionable, pathophysiologically grounded monitoring window for early detection and risk stratification in ischemic cerebrovascular events. In the present study, NLR was significantly higher in the concurrent CI group and identified as a significant predictor of in-hospital first-onset CI, corroborating previous findings and reinforcing the rationale for its inclusion in the predictive model.

Extensive evidence underscores the prognostic significance of carotid plaque characteristics in cardiovascular risk stratification. Jérôme Fichet et al demonstrated 52% of acute coronary syndrome patients exhibit concurrent carotid atherosclerosis,⁴⁹ with ultrasound-derived plaque hardness independently predicting adverse outcomes at 19-month follow-up reported by R Komorovsky.⁵⁰ Systematic reviews further highlight that biomechanical stress parameters, including wall shear stress and plaque wall stress, demonstrate moderate-to-high sensitivity in forecasting plaque-related cerebrovascular events.⁵¹ Longitudinal studies corroborate the superiority of plaque characteristics—such as Total Plaque Risk Score (12-month prognosis⁵²) and carotid intima-media thickness (6.4-year stroke risk⁵³)—in long-term risk prediction. While these plaque-centric biomarkers provide invaluable insights into chronic atherosclerotic progression, their temporal resolution aligns poorly with acute-phase clinical decision-making. Our study strategically focuses on identifying imminent in-hospital CI during AMI management, a critical window where traditional plaque metrics exhibit limited discriminative capacity. Our biomarker panel (CRP, NLR, Killip classification, BUN) detects imminent in-hospital events through distinct pathophysiological pathways - particularly systemic inflammation (CRP/NLR) and acute hemodynamic stress (Killip/BUN). This acute-temporal specificity positions our panel not as a replacement for plaque-based prognostication, but as its necessary counterpart—enabling precision monitoring when anatomical biomarkers lose clinical immediacy.

Limitations

Although this study provides valuable insights into the prediction of CI in AMI patients, several limitations must be acknowledged. First, the study design was a single-center retrospective analysis, which may introduce selection bias and information bias due to the inherent limitations of retrospective data collection. The relatively small sample size further limits the generalizability of our findings. Second, the prediction model was developed and validated using data from a single center, which may limit its external validity. Future multicenter studies with larger, more diverse populations are needed to confirm the predictive efficacy of the model and to develop a more accurate and generalizable tool for clinical application in China. Additionally, the inclusion of only in-hospital CI events may underestimate the true incidence of this complication, as some cases may occur after discharge. Further research should also investigate the potential impact of unmeasured confounders, such as medication adherence and lifestyle factors, on the model’s performance.

Conclusion

This study successfully developed and internally validated the first prediction model for the first onset of CI during hospitalization in AMI patients, using five routinely available clinical indicators: Killip classification, PCI treatment, CRP, BUN, and NLR. The model demonstrated robust predictive performance, with an optimal cutoff value of 0.035 determined by ROC curve analysis (sensitivity: 0.783, specificity: 0.711), providing clinicians with a clear intervention threshold. By integrating key domains of cardiac function assessment (Killip classification), hemodynamic intervention (PCI treatment), inflammatory markers (CRP and NLR), and metabolic status (BUN), the model enables accurate quantification of individualized risk probabilities. This facilitates early identification and stratified management of high-risk populations, providing an evidence-based foundation for prophylactic antithrombotic therapy and multidisciplinary care. Future research should focus on external validation through multicenter prospective cohort studies and further optimization using advanced machine learning algorithms to improve dynamic prediction capabilities. Such efforts will facilitate the integration of this model into clinical decision support systems, ultimately improving patient outcomes in this high-risk population.