Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2025 Oct 16;8(1):613.
doi: 10.1038/s41746-025-01983-7.

External validation of PreOpNet to predict 30-day mortality after major non-cardiac surgery using digital electrocardiogram

Arnaud Champetier #  1   2 , Pedro Lopez-Ayala #  1   2 , Christian Puelacher  1   3 , Koray Durak  1 , Emel Kaplan  1   2 , Danielle Gualandro  1   2 , Noemi Glarner  1   2 , Mirjam Pargger  1   2 , Gabrielle Huré  1   2 , Katrin Burri-Winkler  1   4 , Daniel Bolliger  4 , Luzius A Steiner  4   5 , Stefan Schären  6 , Edin Mujagic  7 , Martin Clauss  8   9 , Andreas M Mueller  8 , Didier Lardinois  10 , Jasper Boeddinghaus  1 , Felix Mahfoud  1 , Ivo Strebel  11 , Christian Mueller  12   13
Affiliations

External validation of PreOpNet to predict 30-day mortality after major non-cardiac surgery using digital electrocardiogram

Arnaud Champetier et al. NPJ Digit Med. .

Abstract

PreOpNet is a novel deep-learning algorithm using 12-lead digital electrocardiogram (ECG) for preoperative risk assessment of all-cause death and major adverse cardiac events (MACE) within 30 days. Its performance in European high-risk patients undergoing major non-cardiac surgery-the target population for guideline-recommended risk assessment-and comparison to high-sensitivity cardiac troponin T (hs-cTnT), is unknown. In a prospective European study (2014-2019), 6098 high-risk patients with available ECGs were enrolled. PreOpNet showed moderate discrimination for death (AUC 0.707) and MACE (0.675), but overestimated risk. It outperformed the revised cardiac risk index (RCRI) for death (AUC 0.644), but not for MACE (0.662). Hs-cTnT remained superior for both outcomes (AUC 0.762 and 0.743). Importantly, PreOpNet provided incremental prognostic value when combined with RCRI and/or hs-cTnT. PreOpNet has limited benefit for preoperative risk stratification in high-risk surgical patients as a stand-alone test. However, it holds promise when used in conjunction with RCRI and hs-cTnT. Clinical Trial Registration: ClinicalTrials.gov number: NCT02573532; https://www.clinicaltrials.gov/study/NCT02573532 .

PubMed Disclaimer

Conflict of interest statement

Competing interests: P.L.A. has received research grants from the Swiss Heart Foundation (FF20079, FF21103 and FF24149) and speaker’s honoraria from Quidel, paid to the institution, Roche Diagnostics and Polymedco in the last 36 months, all outside the submitted work. J.B. is supported by an Edinburgh Doctoral College Scholarship and research grants from the University of Basel, the University Hospital of Basel, the Division of Internal Medicine, the Swiss Academy of Medical Sciences, the Gottfried and Julia Bangerter-Rhyner Foundation, the Swiss National Science Foundation, the Swiss Heart Foundation, and has received honoraria from Siemens, Roche Diagnostics, Ortho Clinical Diagnostics, Quidel Corporation, and Beckman Coulter, and travel support from Medtron-ic and Vascularmedical, all outside the submitted work. C.M. reports receiving research support from the the Swiss National Science Foundation, the Swiss Heart Foundation, the University Hos-pital Basel, the University of Basel, Abbott, Astra Zeneca, Boehringer Ingelheim, Beckman Coul-ter, BRAHMS, Idorsia, Novartis, Ortho Clinical, Quidel, Roche, Siemens, SpinChip, Upstream, and Sphingotec, as well as speaker/consulting honoraria from Acon, Amgen, Astra Zeneca, Bayer, Boehringer Ingelheim, Daiichi Sankyo, Idorsia, Osler, Novartis, Novo Nordisk, Roche, SpinChip, and Sanofi, all paid to the institution. F.M. has been supported by Deutsche Forschungsgemein-schaft (SFB TRR219, Project-ID 322900939) and Deutsche Herzstiftung. Saarland University has received scientific support from Ablative Solutions, Medtronic and ReCor Medical. Until May 2024, F.M. has received speaker honoraria/consulting fees from Ablative Solutions, Astra-Zeneca, Inari, Medtronic, Merck, Novartis, Philips and ReCor Medical. C.P. received research support from Roche Diagnostics and the Swiss Heart Foundation, as well as chaired an advisory board with honoraria from Roche Diagnostics paid to the institution. E.K. reports support from the Swiss Heart Foundation, the University Hospital Basel, Bangerter-Rhyner Foundation, and speaking/consulting fees from SpinChip, Boehringer Ingelheim. All other authors declare no competing interests.

Figures

Fig. 1
Fig. 1. Patient flowchart.
ECG electrocardiogram, hs-cTnT high-sensitivity cardiac troponin T.
Fig. 2
Fig. 2. Prognostic performance of the PreOpNet model.
a Distribution of predicted risk for death predictions; b Receiver-operating-characteristic curve showing the discrimination of PreOpNet for predicting death within 30 days after surgery; c Calibration curve for PreOpNet, assessing the agreement between predicted and observed risk of death within 30 days after surgery. The red line represents optimal calibration, the black solid line the calibration curve, and the black dashed line the 95% confidence interval of the calibration curve; d Distribution of predicted risk for MACE predictions; e Receiver-operating-characteristic curve showing the discrimination of PreOpNet for predicting MACE within 30 days after surgery; f Calibration curve for PreOpNet, assessing the agreement between predicted and observed risk of MACE within 30 days after surgery. The red line represents optimal calibration, the black solid line the calibration curve, and the black dashed line the 95% confidence interval of the calibration curve. AUC Area under the receiver operating characteristic curve, CI Confidence interval.
Fig. 3
Fig. 3. Forest plot of the discrimination of the PreOpNet model across patient subgroups.
Discrimination for predicting a death and b MACE within 30 days. AUC Area under the receiver operating characteristic curve, CI confidence interval, ECG electrocardiogram, RCRI Revised cardiac risk index, eGFR Estimated glomerular filtration rate, Hs-cTnT High-sensitivity cardiac troponin T.
Fig. 4
Fig. 4. Discrimination of PreOpNet, hs-cTnT and RCRI.
Receiver-operating-characteristic curves showing the discrimination of the PreOpNet model (green), high-sensitivity cardiac troponin T (blue), and the RCRI (red) for predicting a death and b MACE within 30 days. RCRI Revised cardiac risk index, AUC Area under the receiver operating characteristic curve, CI Confidence interval, Hs-cTnT High-sensitivity cardiac troponin T.
Fig. 5
Fig. 5. Prognostic performance of PreOpNet using a population-wise normalization as the final step of preprocessing.
a Distribution of predicted risk for death predictions; b Receiver-operating-characteristic curve showing the discrimination of PreOpNet, hs-cTnT and RCRI for predicting death within 30 days after surgery; c Calibration curve of PreOpNet assessing the agreement between predicted and observed risk of death within 30 days after surgery. The red line represents optimal calibration, the black solid line the calibration curve, and the black dashed line the 95% confidence interval of the calibration curve; d Distribution of predicted risk for MACE predictions; e Receiver-operating-characteristic curve showing the discrimination of PreOpNet, hs-cTnT and RCRI for predicting MACE within 30 days after surgery; f Calibration curve of PreOpNet assessing the agreement between predicted and observed risk of MACE within 30 days after surgery. The red line represents optimal calibration, the black solid line the calibration curve, and the black dashed line the 95% confidence interval of the calibration curve. RCRI Revised cardiac risk index, AUC Area under the receiver operating characteristic curve, CI Confidence interval, Hs-cTnT High-sensitivity cardiac troponin T.
See this image and copyright information in PMC

References

    1. Nepogodiev, D. et al. Global burden of postoperative death. Lancet393, 401 (2019). - PubMed
    1. Lee, T. H. et al. Derivation and prospective validation of a simple index for prediction of cardiac risk of major noncardiac surgery. Circulation100, 1043–1049 (1999). - PubMed
    1. Halvorsen, S. et al. 2022 ESC Guidelines on cardiovascular assessment and management of patients undergoing non-cardiac surgery: developed by the task force for cardiovascular assessment and management of patients undergoing non-cardiac surgery of the European Society of Cardiology (ESC) Endorsed by the European Society of Anaesthesiology and Intensive Care (ESAIC). Eur. Heart J.43, 3826–3924 (2022). - PubMed
    1. Biccard, B. M., Devereaux, P. J. & Rodseth, R. N. Cardiac biomarkers in the prediction of risk in the non-cardiac surgery setting. Anaesthesia69, 484–493 (2014). - PubMed
    1. Ribeiro, A. H. et al. Automatic diagnosis of the 12-lead ECG using a deep neural network. Nat. Commun.11, 1760 (2020). - PMC - PubMed

Associated data