Pre-procedural neutrophil-centred inflammatory biomarkers are independent correlates of post-PCI no-reflow and offer incremental predictive value over conventional clinical variables, with an interpretable LightGBM model yielding optimal discrimination (AUC 0.799) for peri-procedural risk stratification.
No-reflow occurred in 91 of 382 AMI patients (23.8%) undergoing PCI.
Multiple neutrophil-centred indices independently correlated with no-reflow after multivariable adjustment.
The LightGBM machine-learning model yielded the best discrimination among four models tested, with an AUC of 0.799.
Neutrophil-centred inflammatory biomarkers provided significant incremental predictive value over conventional clinical variables.
Neutrophil count, NLR, and MHR were associated with no-reflow in STEMI but not NSTEMI patients, though formal interaction tests found no significant AMI-subtype effect modification.
LASSO regression was used for predictor selection, and the final seven-predictor model was built with multicollinearity assessment.
The authors concluded that external validation and recalibration are mandatory before clinical application of the model.
This research studied whether routine blood tests measuring inflammation — particularly markers related to neutrophils (a type of white blood cell) — can predict a dangerous complication called 'no-reflow' in heart attack patients. No-reflow occurs when a blocked coronary artery is physically opened by a procedure (PCI/stenting), but blood still doesn't flow properly to the heart muscle, which worsens outcomes. In a group of 382 heart attack patients, about 1 in 4 (23.8%) experienced no-reflow. The study found that several neutrophil-based inflammation scores measured before the procedure were independently linked to no-reflow risk, even after accounting for other known risk factors. The researchers built and compared four artificial intelligence (machine learning) models to predict no-reflow, with the best-performing model (LightGBM) achieving an AUC (accuracy measure) of about 0.80 — meaning it could correctly distinguish patients who would and wouldn't develop no-reflow roughly 80% of the time. They also used a technique called SHAP to make the model's decisions transparent and explainable, showing that the type of heart attack (STEMI vs. NSTEMI) and neutrophil-related markers were the most influential factors. Notably, the inflammatory markers added meaningful predictive information beyond what standard clinical variables alone could provide. This research suggests that simple, widely available blood inflammation tests — particularly those involving neutrophil counts and ratios — could help doctors better identify which heart attack patients are at highest risk for no-reflow before performing the procedure, potentially allowing for more tailored treatment strategies. However, the model was only tested on patients from a single center, and the authors emphasize that further testing in different patient populations is required before this tool could be used in routine clinical practice.
Liu R, Peng Y, Qin H, Zhou Y, Li X, Wang H, et al.. (2026). Interpretable machine learning analysis of immune-inflammatory biomarkers for predicting no-reflow in patients with acute myocardial infarction after percutaneous coronary intervention.. Annals of medicine. https://doi.org/10.1080/07853890.2026.2747562