Methods

This single-centre retrospective cohort study was conducted at Erasmus University Medical Centre, a tertiary university medical centre in Rotterdam, The Netherlands. The study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines and approved by the local research ethics committee, which waived the requirement for informed consent (MEC-2023-0552; approved 19 October 2023). The study was registered in the Overview of Medical Research in The Netherlands (NL-011210). Data were pseudonymised and stored in a secure digital research environment.

Study population

Adult patients (≥18 yr) undergoing noncardiac surgery between 2017 and 2022 were included if at least one postoperative high-sensitivity cardiac troponin T (hs-cTnT) concentration exceeded 50 ng L−1 within 72 h of surgery. The sample size was based on the available data. In patients with multiple procedures, only the first qualifying operation was analysed. Cardiac surgery, lung transplantation, and day-case surgery were excluded.

Troponin measurements

Postoperative hs-cTnT was measured routinely after arterial vascular surgery and at the enhanced perioperative care (EPC) unit, a postoperative high-care unit with overnight monitoring for patients at elevated perioperative risk. These groups were considered to be high-risk patients, and therefore, routine postoperative troponin measurement in these settings had been adopted as institutional clinical practice, intended to reflect a pragmatic application of the 2014 European Society of Cardiology (ESC) guideline recommendation applicable during the study period.13 Outside these settings, testing was performed at the discretion of the treating clinician. Preoperative troponin measurements were only available in a subset of patients and were therefore not taken into account. Measurements were performed using the Roche Elecsys® hs-cTnT assay (upper reference limit 14 ng L−1).

Data collection

Baseline characteristics were acquired from medical records and consisted of age, sex, type of surgery, and a previous medical history of hypertension, diabetes mellitus with or without insulin use, coronary artery disease, previous myocardial infarction, congestive heart failure, chronic kidney disease, cerebrovascular disease, atrial fibrillation, moderate or severe valvular disease, chronic obstructive pulmonary disease, and peripheral artery disease. Surgery was classified as elective when scheduled in advance; all other procedures, including semi-emergency and emergency operations, were classified as non-elective. Additional perioperative laboratory measurements and admission details were extracted from the institution’s electronic medical record storage database.

Hierarchical classification of myocardial injury

After data extraction, the aetiology of myocardial injury was hierarchically adjudicated based on the patient's context at the time of troponin threshold breach, as documented in the electronic health record, and registered in Castor EDC (Castor Electronic Data Capture, Amsterdam, The Netherlands) using a prespecified hierarchical framework adapted from a previously published classification model (Fig. 1).12,14 In accordance with the original framework, each patient was assigned to one mutually exclusive category representing the single most plausible dominant mechanism, classified as either extra-cardiac or cardiac. Extra-cardiac causes included sepsis, stroke, pulmonary embolism, acute or chronic renal failure, cardiothoracic trauma, and other extra-cardiac causes.15 If no extra-cardiac causes were identified, specific cardiac causes were assessed. Cardiac causes comprised type 1 myocardial infarction, non-sinus tachyarrhythmia, or acute heart failure. In the absence of these conditions, myocardial injury was classified as type 2 myocardial injury and further categorised according to the presence or absence of a possible identifiable trigger. Triggers for type 2 myocardial injury included hypotension, hypoxaemia, anaemia, or sinus tachycardia. Patients without an identifiable trigger were classified as having type 2 myocardial injury without a documented precipitant. Applied definitions are presented in Supplement 1.

Fig 1: Hierarchical adjudication protocol.

Classification and reassessment

The most likely aetiology of myocardial injury was assessed by two independent consultant anaesthetists using the prespecified algorithm described earlier (Fig. 1). Blinded reassessment was performed by two trained anaesthesia-affiliated healthcare professionals using the same protocol. Inter-rater agreement was quantified, and discrepant classifications were retained without consensus resolution and visualised using a Sankey diagram.

Outcome

The primary outcome was the aetiological distribution of myocardial injury and its classification as extra-cardiac or cardiac. Secondary outcomes included inter-rater agreement after blinded reassessment of myocardial injury aetiology and in-hospital mortality.

Statistical analysis

All statistical analyses were performed using R version 4.5.2 (R Foundation for Statistical Computing, Vienna, Austria) and RStudio version 2025.09.1 (Posit Software, PBC, Boston, MA, USA). Baseline characteristics of the overall cohort and comparisons between extra-cardiac and cardiac myocardial injury were summarised as counts (%) for categorical variables and compared using Pearson’s χ2 test. Continuous variables were assessed for normality visually and using the Shapiro–Wilk test. Normally distributed variables were summarised as mean (standard deviation) and compared using the independent samples t-test, whereas non-normally distributed variables were summarised as median [interquartile range] and compared using the Mann–Whitney U-test. Two sensitivity analyses were performed: one excluding patients with pre-existing chronic kidney disease and one restricted to hs-cTnT >50 ng L−1 measured at the EPC unit or after vascular surgery. Inter-rater agreement for binary myocardial injury classification (extra-cardiac and cardiac) between consultant anaesthetists and anaesthesia-affiliated healthcare professionals was quantified using percentage agreement and Cohen’s κ with 95% confidence interval (CI). In-hospital death and discharge alive within 30 days were analysed as competing events using cumulative incidence functions, with between-group differences assessed using Gray’s test. Separate multivariable logistic regression analyses were performed for extra-cardiac and cardiac myocardial injury, with in-hospital mortality as the dependent variable. Covariates were emergency surgery, Revised Cardiac Risk Index, and baseline characteristics not included in the Revised Cardiac Risk Index.16 P<0.05 was considered statistically significant.