Article Text

Download PDFPDF

Machine learning to support decision-making for cardiac surgery during the acute phase of infective endocarditis
  1. Erwan Donal1,
  2. Erwan Flecher2,
  3. Pierre Tattevin3
  1. 1 Department of Cardiology, CHU Rennes, Rennes, France
  2. 2 Department of Cardiac Surgery, CHU Pontchaillou, Rennes, France
  3. 3 Department for infectious diseases, CHU Rennes – Hôpital pontchaillou, Rennes, France
  1. Correspondence to Dr Erwan Donal, Department of Cardiology, CHU Rennes, CHU Pontchaillou, Rennes 35033, France; erwan.donal{at}chu-rennes.fr

Statistics from Altmetric.com

Rapid identification of patients at high risk of death may trigger additional therapeutic interventions, which in turn may change the course of the disease and improve prognosis. For infective endocarditis (IE), inhospital mortality rate remains high—from 15% to 30%—despite major achievements over the last 20 years in diagnostic tools (imaging, microbiology) and cardiac surgery.1 Prognostic factors include, in addition to the usual suspects (eg, age, comorbidities, heart failure), variables specific to the pathophysiology of IE: periannular complication, Staphylococcus aureus infection or persistence of positive blood cultures >72 hours after initiation of appropriate antibiotic treatment. The recommendation is to early transfer these patients to referral centres, where an endocarditis team will adjust their management and evaluate the indication of, and—in selected cases—the best timing for, cardiac surgery.2

The endocarditis team is a great innovation strongly supported by recent guidelines from America and Europe: The American 2015 guidelines3 stated that ‘Decisions on the indication and timing of surgical intervention should be determined by a multispecialty team with expertise in cardiology, imaging, cardiothoracic surgery, and infectious diseases’, while the European guidelines went further, with a specific ‘endocarditis team’ section, where the desirable characteristics of the endocarditis team are comprehensively elaborated, as well as their role throughout the process. Although not all patients with IE will require cardiac surgery during the acute phase, this will eventually be the case for approximately half of them, and the endocarditis team should be rapidly contacted for the decision-making process.1 Risk stratification plays an important role for the selection of cases that will benefit from cardiac surgery, and the reasoning is based on this risk–benefit balance, that is, what is the risk associated with an early surgical approach compared with a conservative management. In addition, timing of cardiac surgery also plays an important role in the estimation of this risk–benefit balance. For instance, early surgery should be considered when a large vegetation is identified, for the prevention of embolism.4 However, as the risk of embolism decreases dramatically following the start of appropriate antibiotics, the benefit of cardiac surgery will be lost after a few days, while the risks associated with the procedure will remain grossly similar. Hence, the risk:benefit ratio supports the indication of cardiac surgery for the prevention of embolism only if performed during the first days of medical treatment.5

Carmen Olmos et al 6 aimed to develop a simple bedside prediction score, the RISK-E score, to assist clinicians and the endocarditis team in the difficult decision-making process for cardiac surgery during the acute phase of IE. The RISK-E score, simple and readily available, may be a valuable tool for decision-making, in line with scores that are currently advocated for the risk assessment in surgery for other cardiac diseases, to assist the selection of investigations for the diagnosis of suspected pulmonary embolism, or the prescription of anticoagulant for atrial arrhythmias. The RISK-E score is calculated through the addition of different simple parameters including age (≤52 years old, 0 point; 53–63 years old, 9 points; 64–73 years old, 13 points; and ≥74 years old, 14 points), prosthetic valve IE (6 points), periannular complications (5 points), S. aureus or fungal IE (9 points), acute renal failure (5 points), septic shock (7 points), cardiogenic shock (15 points) and thrombocytopaenia (7 points). The rigorous methodology, the large number of patients enrolled, the multicentric design and the construction of three different cohorts (ie, derivation and validation cohorts within a Spanish network of reference centres, plus an external validation within a French prospective cohort) are all important assets. This innovative RISK-E score performed significantly better than four existing surgical scores, EuroSCORE I and II (European System for Cardiac Operative Risk Evaluation), PALSUSE (Prosthesis, Age ≥70 years, Large intracardiac destruction, Staphylococcus, Urgent surgery, sex (female), EuroSCORE ≥10) and STS-IE (Society of Thoracic Surgeons’s Infective endocarditis score), not only in the derivation score—as is the case for most scores—but also in the validation Spanish cohort, which suggests that it may represent a significant progress in the field, with a better accuracy to predict mortality after cardiac surgery during the acute phase of IE.

Although the authors are to be commended for this landmark study and the robustness of their methodology, two significant limitations not mentioned in their manuscript must be outlined. First, the cohort was built from a database of IE cases enrolled during almost 20 years (1996–2014), while various changes probably occurred in the management of IE in the participating centres. Significant changes in the composition of medical teams or the use of local protocols most likely occurred, while dramatic progress have been achieved in imaging studies, molecular biology, intensive care and cardiac surgery, as documented by changes in international guidelines for IE over the last 20 years. These changes most likely translated into heterogeneity in the management of patients with IE over the study period. Second, and most importantly, the RISK-E score did not perform as well in the external validation cohort (ie, The 2010–2015 French monocentric cohort), with an area under the receiver operating characteristic (ROC) curve of 0.76 (95% CI: 0.64 to 0.88), quite similar to the performance of the logistic EuroSCORE, EuroSCORE II and STS-IE score in the Spanish multicentre cohort, with areas under the ROC curve of, respectively, 0.76 (95% CI 0.71 to 0.82), 0.76 (95% CI 0.70 to 0.82) and 0.74 (95% CI 0.68 to 0.79). This loss of performance in another cohort raises doubts on the validity of the RISK-E score outside of the cohorts used for its construction.

The predictive accuracy of this novel surgical risk score performed well, but it will not, in our opinion, replace the necessary expertise of the endocarditis team. In line with the scores used in the decision-making process for trans aortic valve implantation (TAVI), the RISK-E score may be used as a first filter that could help to not underestimate or overestimate the risk of death in an individual patient, especially outside of tertiary care centres.

References

Footnotes

  • Competing interests None declared.

  • Provenance and peer review Commissioned; internally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • Valvular heart disease
    Carmen Olmos Isidre Vilacosta Gilbert Habib Luis Maroto Cristina Fernández Javier López Cristina Sarriá Erwan Salaun Salvatore Di Stefano Manuel Carnero Sandrine Hubert Carlos Ferrera Gabriela Tirado Afonso Freitas-Ferraz Carmen Sáez Javier Cobiella Juan Bustamante-Munguira Cristina Sánchez-Enrique Pablo Elpidio García-Granja Cecile Lavoute Benjamin Obadia David Vivas Ángela Gutiérrez José Alberto San Román