A machine-learning-based model for predicting 90-day postoperative mortality in esophageal cancer surgery

Objective: to develop a preoperative model for predicting 90-day mortality to optimize treatment strategy in patients with resectable esophageal cancer and Siewert type I gastroesophageal junction cancer.

Material and Methods. A retrospective cohort study included 225 patients with resectable esophageal cancer (n=179) and Siewert type I gastroesophageal junction cancer (n=46). A total of 79 preoperative variables were analyzed. Risk factors were selected using univariate and multivariate logistic regression. The dataset was randomly split into training and validation cohorts (70/30). The predictive model was developed using l1-regularized multivariable logistic regression and the TabNet neural network featuring a sparse attentive architecture that enables estimation of feature importance. Discriminative ability and calibration were assessed using AUCROC analysis and the Hosmer–Lemeshow test.

Results. The in-hospital mortality, 30-day mortality and 90-day mortality rates were 8.9 % (n=20), 1.3 % (n=3), and 6.2 % (n=14), respectively. The most significant predictors of mortality included preoperative weight loss >10 %, grade 3–4 dysphagia, ECOG performance status ≥2, COPD, preoperative dyspnea, history of major thoracic surgery, TNM stage N2, tumor proximity to or invasion of the trachea/main bronchi, cerebrovascular disease history, diabetes mellitus, and complications during or after radiotherapy. The logistic regression model demonstrated an AUC-ROC of 0.86 (95 % Ci 0.79–0.92) in the training cohort and 0.77 (95 % CI 0.72–0.83) in the validation cohort. Calibration was supported by non-significant Hosmer–Lemeshow χ² tests in both cohorts: χ²=1.98, df=8 (p=0.98) and χ²=4.72, df=8 (p=0.78), respectively. The TabNet neural network achieved improved discrimination with AUC-ROC values of 0.95 (95 % CI 0.92–0.98) in the training cohort and 0.86 (95 % Ci 0.82–0.96) in the validation cohort; calibration results were χ²=8.38, df=8 (p=0.39) and χ²=9.46, df=8 (p=0.30), respectively. The model has been implemented as a software application and can be used in real-world clinical scenarios.

Conclusion. In this study, a machine learning-based model for preoperative 90-day mortality prediction was developed. It demonstrated good validity and high discriminative performance. The findings support the promise of machine learning methods for clinical risk prediction in cancer surgery and highlight their potential for implementation in routine healthcare practice.

1. Low D.E., Alderson D., Cecconello I., Chang A.C., Darling G.E., DʼJourno X.B., Griffin S.M., Hölscher A.H., Hofstetter W.L., Jobe B.A., Kitagawa Y., Kucharczuk J.C., Law S.Y., Lerut T.E., Maynard N., Pera M., Peters J.H., Pramesh C.S., Reynolds J.V., Smithers B.M., van Lanschot J.J. International Consensus on Standardization of Data Collection for Complications Associated With Esophagectomy: Esophagectomy Complications Consensus Group (ECCG). Ann Surg. 2015; 262(2): 286–94. doi: 10.1097/sLA.0000000000001098.

2. Brusselaers N., Mattsson F., Lagergren J. Hospital and surgeon volume in relation to long-term survival after oesophagectomy: systematic review and meta-analysis. Gut. 2014; 63(9): 1393–400. doi: 10.1136/gutjnl-2013-306074.

3. Talsma A.K., Lingsma H.F., Steyerberg E.W., Wijnhoven B.P., van Lanschot J.J. The 30-day versus in-hospital and 90-day mortality after esophagectomy as indicators for quality of care. Ann Surg. 2014; 260(2): 267–73. doi: 10.1097/SLA.0000000000000482.

4. Luketich J.D., Pennathur A., Awais O., Levy R.M., Keeley S., Shende M., Christie N.A., Weksler B., Landreneau R.J., Abbas G., Schuchert M.J., Nason K.S. Outcomes after minimally invasive esophagectomy: review of over 1000 patients. Ann Surg. 2012; 256(1): 95–103. doi: 10.1097/SLA.0b013e3182590603.

5. Sjoquist K.M., Burmeister B.H., Smithers B.M., Zalcberg J.R., Simes R.J., Barbour A., Gebski V.; Australasian Gastro-Intestinal Trials Group. Survival after neoadjuvant chemotherapy or chemoradiotherapy for resectable oesophageal carcinoma: an updated meta-analysis. Lancet Oncol. 2011; 12(7): 681–92. doi: 10.1016/S1470-2045(11)70142-5.

6. Low D.E., Kuppusamy M.K., Alderson D., Cecconello I., Chang A.C., Darling G., Davies A., D’Journo X.B., Gisbertz S.S., Griffin S.M., Hardwick R., Hoelscher A., Hofstetter W., Jobe B., Kitagawa Y., Law S., Mariette C., Maynard N., Morse C.R., Nafteux P., Pera M., Pramesh C.S., Puig S., Reynolds J.V., Schroeder W., Smithers M., Wijnhoven B.P.L. Benchmarking Complications Associated with Esophagectomy. Ann Surg. 2019; 269(2): 291–98. doi: 10.1097/SLA.0000000000002611.

7. D’Journo X.B., Berbis J., Jougon J., Brichon P.Y., Mouroux J., Tiffet O., Bernard A., de Dominicis F., Massard G., Falcoz P.E., Thomas P., Dahan M. External validation of a risk score in the prediction of the mortality after esophagectomy for cancer. Dis Esophagus. 2017; 30(1): 1–8. doi: 10.1111/dote.12447.

8. Takeuchi H., Miyata H., Gotoh M., Kitagawa Y., Baba H., Kimura W., Tomita N., Nakagoe T., Shimada M., Sugihara K., Mori M. A risk model for esophagectomy using data of 5354 patients included in a Japanese nationwide web-based database. Ann Surg. 2014; 260(2): 259–66. doi: 10.1097/SLA.0000000000000644.

9. Fuchs H.F., Harnsberger C.R., Broderick R.C., Chang D.C., Sandler B.J., Jacobsen G.R., Bouvet M., Horgan S. Simple preoperative risk scale accurately predicts perioperative mortality following esophagectomy for malignancy. Dis Esophagus. 2017; 30(1): 1–6. doi: 10.1111/dote.12451.

10. Arik S.Ö., Pfister T. TabNet: Attentive Interpretable Tabular Learning. Proceedings of the AAAI Conference on Artificial Intelligence. 2021; 35(8): 6679–87. https://doi.org/10.1609/aaai.v35i8.16826.

11. Brown J., Lewis W.G., Foliaki A., Clark G.W.B., Blackshaw G.R.J.C., Chan D.S.Y. Colonic Interposition After Adult Oesophagectomy: Systematic Review and Meta-analysis of Conduit Choice and Outcome. J Gastrointest Surg. 2018; 22(6): 1104–11. doi: 10.1007/s11605-018-3735-8.

12. Kim J.H., Yun J.K., Kim C.W., Kim H.R., Kim Y.H. Long-Term Outcomes of Colon Conduits in Surgery for Primary Esophageal Cancer: A Propensity Score-Matched Comparison to Gastric Conduits. J Chest Surg. 2024; 57(1): 53–61. doi: 10.5090/jcs.23.074.

13. Oshikiri T., Takiguchi G., Hasegawa H., Yamamoto M., Kanaji S., Yamashita K., Matsuda T., Nakamura T., Suzuki S., Kakeji Y. Postoperative recurrent laryngeal nerve palsy is associated with pneumonia in minimally invasive esophagectomy for esophageal cancer. Surg Endosc. 2021; 35(2): 837–44. doi: 10.1007/s00464-020-07455-1.

14. Koyanagi K., Igaki H., Iwabu J., Ochiai H., Tachimori Y. Recurrent Laryngeal Nerve Paralysis after Esophagectomy: Respiratory Complications and Role of Nerve Reconstruction. Tohoku J Exp Med. 2015; 237(1): 1–8. doi: 10.1620/tjem.237.1.

15. Scholtemeijer M.G., Seesing M.F.J., Brenkman H.J.F., Janssen L.M., van Hillegersberg R., Ruurda J.P. Recurrent laryngeal nerve injury after esophagectomy for esophageal cancer: incidence, management, and impact on shortand long-term outcomes. J Thorac Dis. 2017; 9(s8): 868–78. doi: 10.21037/jtd.2017.06.92.

16. Wu W., Zhang Z., Xu Z., Zhang L., Jiang J., Lin F. A systematic review and meta-analysis of intraoperative neuromonitoring (IONM) of the recurrent laryngeal nerve during minimally invasive esophagectomy. J Thorac Dis. 2024; 16(12): 8550–64. doi: 10.21037/jtd-24-1024.

17. Yuda M., Nishikawa K., Ishikawa Y., Takahashi K., Kurogochi T., Tanaka Y., Matsumoto A., Tanishima Y., Mitsumori N., Ikegami T. Intraoperative nerve monitoring during esophagectomy reduces the risk of recurrent laryngeal nerve palsy. Surg Endosc. 2022; 36(6): 3957–64. doi: 10.1007/s00464-021-08716-3.

18. Rice T.W., Ishwaran H., Ferguson M.K., Blackstone E.H., Goldstraw P. Cancer of the Esophagus and Esophagogastric Junction: An Eighth Edition Staging Primer. J Thorac Oncol. 2017; 12(1): 36–42. doi: 10.1016/j.jtho.2016.10.016.

19. Lagergren J., Mattsson F., Zylstra J., Chang F., Gossage J., Mason R., Lagergren P., Davies A. Extent of Lymphadenectomy and Prognosis After Esophageal Cancer Surgery. JAMA Surg. 2016; 151(1): 32–39. doi: 10.1001/jamasurg.2015.2611.