Risk prediction for cardiac surgery and interventions

Abstract

Background: Cardiac surgery and interventions have rapidly evolved and advanced significantly over the last two decades with improving outcomes. Risk models play a critical role in the decision-making for all cardiac procedures. Despite this, they are under-utilised and have not been assessed and validated in New Zealand cardiac surgery cohorts or for predicting morbidities and long-term mortality. There is sparse literature regarding the application of risk models in some clinically important settings such as infective endocarditis surgery and transcatheter aortic valve implantation. Finally the prognostic utility of recently developed high-sensitivity troponin assays in cardiac surgery have not been well evaluated. The aims of this thesis are to address each of these aforementioned issues, in assessing performance of risk scores and troponins in New Zealand cohorts, for predicting mortality and morbidities, and by means of meta-analyses where appropriate. Methods: A literature review of risk modelling and surgical risk scores for cardiac surgery and interventions was conducted. Eight studies were then performed, including six Auckland City Hospital based cohort studies and two meta-analyses. These studies: 1. Compared surgical risk scores for outcomes after isolated coronary artery bypass grafting. 2. Assessed the mortality and morbidities prediction of risk scores for isolated aortic valve replacement 3. Evaluated the utility of risk scores for mitral valve repair and replacement surgery. 4. Compared risk scores for combined aortic valve replacement and coronary bypass grafting surgery 5. Assessed surgical and endocarditis-specific risk scores for infective endocarditis operations. 6. Performed meta-analysis of surgical risk scores for infective endocarditis surgery. 7. Pooled performance of contemporary surgical risk scores when applied to transcatheter aortic valve implantation outcomes. 8. Reviewed the prognostic utility of high-sensitivity troponin T with ECG and/or echocardiographic changes for coronary artery bypass grafting and evaluating the universal definition for type 5 perioperative myocardial infarction. Results: 1. The newer EuroSCORE II, STS and AusSCOREs had improved calibration, but only similar discrimination to EuroSCORE (c-statistic 0.64-0.68) for coronary artery surgery. 2. In isolated aortic valve replacement, all scores had moderate discrimination for operative mortality (cstatistic 0.68-0.75), however the STS Score performed best in the highest surgical risk quintile including for calibration, and also for composite and individual post-operative complications. 3. For isolated mitral valve repair or replacement, all scores had high discrimination (c-statistic 0.82- 0.85) for operative mortality, and the STS Score performed the best for morbidities. 4. In patients undergoing combined aortic valve and coronary surgery, EuroSCORE II and STS Scores had superior discrimination and calibration to EuroSCORE for operative mortality. 5. Endocarditis-specific scores, especially the De Feo-Cotrufo Score performed better than EuroSCOREs at predicting mortality and morbidities after infective endocarditis surgery. 6. Despite our findings above, other studies combined in our meta-analysis found moderate discrimination of EuroSCOREs for predicting operative mortality after infective endocarditis surgery. 7. Surgical risk scores modestly discriminated operative and 1-year mortality with c-statistic 0.62 after transcatheter aortic valve implantation, although EuroSCORE II and STS had better calibration than EuroSCORE which significantly over-estimates risk. 8. Dual criteria of high sensitivity troponin T rise >140ng/L (10 times 99th percentile upper reference limit) with ECG and/or echocardiographic abnormalities, but not other criteria, was independently associated with 30-day and long-term mortality after coronary bypass surgery. Conclusions: Across various types of cardiac surgery, the EuroSCORE, EuroSCORE II and STS scores had similar discrimination, but EuroSCORE significantly over-estimated operative mortality, while the STS Score usually best predicted post-operative complications. Endocarditis-specific scores were superior to EuroSCOREs for endocarditis surgery, while transcatheter aortic valve implantation-specific models and validation are awaited due to modest performance of surgical risk scores in that setting. We also validated the prognostic utility of the Universal Definition's criteria for type 5 myocardial infarction and high-sensitivity troponins after coronary bypass grafting.