Improving 10-year cardiovascular risk prediction in patients with established cardiovascular disease: flexible addition of risk predictors on top of the SMART2 risk score
Year: 2026
Type: article
Abstract: BACKGROUND: Guidelines recommend using the SMART2 model, estimating the risk of recurrent cardiovascular (CV) events, to support treatment decisions in patients with established atherosclerotic CV disease (ASCVD). They further outline that adding biomarkers, comorbidities, anthropometric, and social factors may improve these predictions. AIMS: To investigate the added predictive value of guideline-outlined factors including biomarkers, comorbidities, anthropometric, and social factors on top of the SMART2 model using an approach enabling their use as add-on predictors. METHODS: Patients aged 40-80 with ASCVD were included from 11 cohorts (n=179,382 with 25,789 recurrent CV events). Additional factors included biomarkers (troponin I, NT-proBNP, albuminuria), comorbidities (heart failure, atrial fibrillation, coronary multivessel disease), anthropometric measurements (body-mass index, waist and hip circumference), social (employment, education), and other factors (former smoking, parental CV history). Cross-cohort availability of these factors ranged from 2 cohorts for albuminuria to all cohorts for BMI. These factors were assessed as add-on predictors to the SMART2 model using Fine-Gray models with SMART2 coefficients as offset with recurrent CV events as primary outcome (non-fatal myocardial infarction or stroke, or CV death). Added predictive value was assessed through cohort cross validation by change (Δ) in C--statistic, calibration, and net benefit through decision curve analysis. RESULTS: Sub distribution hazard ratios for additional factors ranged from 0.77 [95% confidence interval 0.75-0.80] for employment status to 1.69 [1.63-1.76] for heart failure history. ΔC-statistic was largest for NT-proBNP (0.0127 [0.0060-0.0193]) and troponin I (0.0100 [0.0020-0.0181]), with statistically significant but smaller ΔC-statistics for employment, heart failure, and atrial fibrillation. Calibration was adequate before and after integration of additional factors. Decision curve analysis demonstrated added net benefit beyond SMART2 for NT-proBNP, heart failure history, atrial fibrillation, albuminuria, current employment, coronary multivessel disease, and education level across clinically relevant thresholds up to 40% predicted risk. CONCLUSIONS: The flexible add-on of guideline-outlined factors on top of SMART2 enables more personalised and improved estimation of recurrent CV event risk in patients with established ASCVD.
Authors: Joris Holtrop, Mari N Gynnild, Matthew Richards, Carl-Emil Lim, Mitja Lääperi, Salil V Deo, Håvard Dalen, Peter Ueda, Sanjay Rajagopalan, Małgorzata Chlabicz, Karol Kaminski, Ignacio Mahíllo-Fernández, Ingvild Saltvedt, Jean-Eudes Dazard, Hanne Ellekjær, Philippe Gabriel Steg, José Tuñón, Torbjørn Omland, Juha Sinisalo, John Munkhaugen, Tomas Jernberg, Deepak L Bhatt, Angela Wood, Frank L J Visseren, Steven H J Hageman, Jannick A N Dorresteijn
Institutions: University Medical Center Utrecht, University of Oslo, Norwegian University of Science and Technology, St Olav's University Hospital, University of Cambridge, National Institute for Health and Care Research, Health Data Research UK, Cambridge Centre for AI in Medicine, Karolinska Institutet, University of Helsinki, Helsinki University Hospital, Louis Stokes Cleveland VA Medical Center, Case Western Reserve University, University of Glasgow, Levanger Hospital, Helse Nord-Trøndelag HF, Medical University of Białystok, Hospital Universitario Fundación Jiménez Díaz, Instituto de Investigación Sanitaria Fundación Jiménez Díaz, Universidad Autónoma de Madrid, Inserm, Université Paris Cité, Assistance Publique – Hôpitaux de Paris, Laboratoire de Recherche Vasculaire Translationnelle, Hôpital Bichat-Claude-Bernard, Akershus University Hospital, Drammen Hospital, Vestre Viken Hospital Trust, Icahn School of Medicine at Mount Sinai
Language: English
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
Estimated APC paid: $3,792
OpenAlex ID: W7128397804