Comparative Performance of HALP, PNI, and CONUT Scores in No-Reflow Among Patients with Acute Coronary Syndrome: A Prospective Study

Background: Nutritional impairment has been associated with adverse outcomes in acute coronary syndrome (ACS), yet its relationship with the no-reflow phenomenon remains incompletely understood. We aimed to compare the performance of the Hemoglobin–Albumin–Lymphocyte–Platelet (HALP), Prognostic Nutritional Index (PNI), and Controlling Nutritional Status (CONUT) scores for no-reflow assessment in patients with ACS. Methods: This prospective single-centre study included 279 consecutive patients with ACS undergoing percutaneous coronary intervention. HALP, PNI, and CONUT scores were calculated from admission laboratory parameters. No-reflow was defined as post-procedural TIMI flow grade ≤ 2 in the absence of mechanical obstruction, with myocardial blush grade used in equivocal cases. Hierarchical logistic regression, receiver operating characteristic (ROC) analysis, net reclassification improvement (NRI), integrated discrimination improvement (IDI), decision curve analysis, and bootstrap validation were performed. Results: No-reflow occurred in 46 patients (16.5%). All three nutritional indices were significantly associated with no-reflow (all p < 0.001). In multivariable analysis, only the CONUT score remained independently associated with no-reflow (OR 1.728, 95% CI 1.226–2.435, p = 0.002). The addition of nutritional indices to the clinical model improved discrimination, increasing the area under the curve from 0.649 to 0.693 for HALP, 0.733 for PNI, and 0.770 for CONUT. CONUT provided the largest likelihood-ratio improvement (χ2 = 25.98, p < 0.001), NRI (0.757, p < 0.001), and IDI (0.104, p < 0.001). Pairwise DeLong comparisons showed no statistically significant differences among the nutritional models. Internal validation of the CONUT model demonstrated good discrimination and calibration (bootstrap-corrected AUC 0.754). Conclusions: Among the evaluated nutritional indices, CONUT showed the largest incremental improvement in model performance; however, statistically significant superiority over HALP and PNI was not demonstrated. These findings should be considered as exploratory and require confirmation in larger multicentre studies.