Inflammation as a missing dimension in early risk stratification after myocardial infarction

Given the heterogeneity of acute coronary syndromes (ACS),1 proper risk stratification after the acute event is essential to support clinical decision-making, including timing of invasive management, tailored use of resources, and personalized secondary prevention measures. The GRACE score has progressively evolved from its original version to GRACE 2.0 and, more recently, to GRACE 3.0, integrating contemporary cohorts and machine learning methods, with particular improvements in female non–ST-segment elevation myocardial infarction (NSTEMI) patients.2–4 However, guiding early invasive management based on the GRACE risk score proved futile,5,6 necessitating novel approaches to risk stratify patient with ACS.7 Inflammation is causally involved in atherogenesis,1,8 and inflammatory biomarkers can refine risk stratification of patients with ACS.9 In this issue of the European Journal of Preventive Cardiology, Sadler et al.10 present the INFLAMmation in Acute Myocardial diseases risk Estimation for Myocardial Infarction (INFLAME-MI) score, an easy-to-use risk model based on routinely available parameters, derived and validated externally in two European cohorts. Most forms of ACS, specifically acute myocardial infarction, are followed by an acute inflammatory burst that is necessary for tissue repair but may become detrimental when excessive or poorly regulated. Several inflammatory parameters, including neutrophil count, neutrophil to lymphocyte ratio (NLR), or elevated level of pro-inflammatory biomarkers, have been consistently associated with worse outcomes after ACS. While neutrophil count, and specifically the NLR, may result from the acute sympathetic (fight-or-flight) response rather than inflammation per se, C-reactive protein is equally important for refined risk prediction, reinforcing a role of inflammation in outcomes after the ACS. Indeed, we have shown previously that the C-reactive protein–relying SEX-SHOCK score markedly improves the prediction of cardiogenic shock following ACS.9 By harnessing elastic net regularization, Sadler et al.10 identified nine variables to be linked to in-hospital mortality, namely, neutrophil count, estimated glomerular filtration rate, C-reactive protein, albumin, troponin, potassium, age, and cardiac arrest at presentation, collectively informing the INFLAME-MI score. For the prediction of in-hospital mortality, INFLAME-MI showed good discriminatory performance in both the derivation and external validation cohorts (AUC 0.88 and 0.92, respectively), though non-superior to that of GRACE. Expectedly, its main strength, however, was calibration. Indeed, in the validation cohort, predicted and observed in-hospital mortality were closely aligned for INFLAME-MI, whereas GRACE consistently overestimated risk. This finding is in line with previous observations,4 likely due to the fact that the initial GRACE score was developed and validated in a cohort in which only ∼1 out of 3 patients underwent percutaneous coronary intervention.2 Importantly, the authors provide an in-depth performance analysis, extending beyond discrimination and calibration. In fact, to gauge its clinical usefulness at predefined decision thresholds, the authors also performed net reclassification improvement and decision curve analyses. Of note, INFLAME-MI provided greater net benefit than GRACE within low-to-intermediate risk groups, which appears to involve the vast majority of ACS patients not requiring immediate revascularization. INFLAME-MI should be viewed in the context of ongoing developments in clinical risk prediction, including the recent evolution of sex-specific and biomarker-enhanced GRACE scores.3,4 While GRACE models provide comprehensive clinical risk assessment, INFLAME-MI offers a simplified, inflammation-focused evaluation based on routinely available laboratory data. These approaches address different dimensions of risk and may be considered complementary in early ACS management. Guiding early invasive management based on the GRACE risk score proved futile in contemporary patients with ACS.5,6 Therefore, we need to re-assess current guideline recommendations and probe novel tools for their clinical utility to risk stratify patients presenting with ACS.7 The INFLAME-MI score is a promising step in this direction, but certainly requires independent validation before routine clinical use can be advocated. Importantly, the model was derived from a single-centre cohort and externally validated in a relatively small patient population, supporting the need for larger, prospective, multicentre studies. Indeed, the very low number of events in the validation cohort (n = 32) confers a substantial risk of model overfit, a concern further underscored by the reported wide margins of error. Moreover, ‘absence of evidence is not evidence of absence’, a notion put forth by Altman and Bland decades ago.11 Accordingly, although the INFLAME-MI score did not demonstrate statistical superiority in terms of model discrimination (P = 0.256), a superiority-based analysis does not justify concluding that its discriminatory performance is comparable or event equivalent to that of GRACE. Finally, as previously observed with GRACE, improved risk prediction does not necessarily translate into changes in clinical management or outcomes in prospectively designed randomized controlled trials,5,6 and future studies should assess the impact of INFLAME-MI on decision-making in real-world practice. In summary, Sadler et al.10 present a biologically coherent and innovative risk score that integrates inflammatory biomarkers into early prognostic assessment after ACS. By showing good calibration and improved clinical utility within relevant decision thresholds, INFLAME-MI may add a meaningful biological dimension to existing risk stratification strategies to complement contemporary risk models as risk assessment continues to evolve. Simon Kraler (Conceptualization [equal], Writing—original draft [equal], Writing—review & editing [equal]), Yifan Wang (Conceptualization [equal], Writing—original draft [equal], Writing—review & editing [equal]), and Luca Liberale (Conceptualization [lead], Writing—original draft [lead], Writing—review & editing [lead]) No new data were generated or analysed in support of this research.