Machine Learning Risk Prediction Model of 90-day Mortality After Gastrectomy for Cancer.
Journal
Annals of surgery
ISSN: 1528-1140
Titre abrégé: Ann Surg
Pays: United States
ID NLM: 0372354
Informations de publication
Date de publication:
01 11 2022
01 11 2022
Historique:
pubmed:
23
7
2022
medline:
12
10
2022
entrez:
22
7
2022
Statut:
ppublish
Résumé
To develop and validate a risk prediction model of 90-day mortality (90DM) using machine learning in a large multicenter cohort of patients undergoing gastric cancer resection with curative intent. The 90DM rate after gastrectomy for cancer is a quality of care indicator in surgical oncology. There is a lack of well-validated instruments for personalized prognosis of gastric cancer. Consecutive patients with gastric adenocarcinoma who underwent potentially curative gastrectomy between 2014 and 2021 registered in the Spanish EURECCA Esophagogastric Cancer Registry database were included. The 90DM for all causes was the study outcome. Preoperative clinical characteristics were tested in four 90DM predictive models: Cross Validated Elastic regularized logistic regression method (cv-Enet), boosting linear regression (glmboost), random forest, and an ensemble model. Performance was evaluated using the area under the curve by 10-fold cross-validation. A total of 3182 and 260 patients from 39 institutions in 6 regions were included in the development and validation cohorts, respectively. The 90DM rate was 5.6% and 6.2%, respectively. The random forest model showed the best discrimination capacity with a validated area under the curve of 0.844 [95% confidence interval (CI): 0.841-0.848] as compared with cv-Enet (0.796, 95% CI: 0.784-0.808), glmboost (0.797, 95% CI: 0.785-0.809), and ensemble model (0.847, 95% CI: 0.836-0.858) in the development cohort. Similar discriminative capacity was observed in the validation cohort. A robust clinical model for predicting the risk of 90DM after surgery of gastric cancer was developed. Its use may aid patients and surgeons in making informed decisions.
Sections du résumé
OBJECTIVE
To develop and validate a risk prediction model of 90-day mortality (90DM) using machine learning in a large multicenter cohort of patients undergoing gastric cancer resection with curative intent.
BACKGROUND
The 90DM rate after gastrectomy for cancer is a quality of care indicator in surgical oncology. There is a lack of well-validated instruments for personalized prognosis of gastric cancer.
METHODS
Consecutive patients with gastric adenocarcinoma who underwent potentially curative gastrectomy between 2014 and 2021 registered in the Spanish EURECCA Esophagogastric Cancer Registry database were included. The 90DM for all causes was the study outcome. Preoperative clinical characteristics were tested in four 90DM predictive models: Cross Validated Elastic regularized logistic regression method (cv-Enet), boosting linear regression (glmboost), random forest, and an ensemble model. Performance was evaluated using the area under the curve by 10-fold cross-validation.
RESULTS
A total of 3182 and 260 patients from 39 institutions in 6 regions were included in the development and validation cohorts, respectively. The 90DM rate was 5.6% and 6.2%, respectively. The random forest model showed the best discrimination capacity with a validated area under the curve of 0.844 [95% confidence interval (CI): 0.841-0.848] as compared with cv-Enet (0.796, 95% CI: 0.784-0.808), glmboost (0.797, 95% CI: 0.785-0.809), and ensemble model (0.847, 95% CI: 0.836-0.858) in the development cohort. Similar discriminative capacity was observed in the validation cohort.
CONCLUSIONS
A robust clinical model for predicting the risk of 90DM after surgery of gastric cancer was developed. Its use may aid patients and surgeons in making informed decisions.
Identifiants
pubmed: 35866643
doi: 10.1097/SLA.0000000000005616
pii: 00000658-202211000-00006
doi:
Types de publication
Journal Article
Multicenter Study
Langues
eng
Sous-ensembles de citation
IM
Pagination
776-783Informations de copyright
Copyright © 2022 Wolters Kluwer Health, Inc. All rights reserved.
Déclaration de conflit d'intérêts
The authors report no conflicts of interest.
Références
De Manzoni G, Marrelli D, Baiocchi GL, et al. The Italian Research Group for Gastric Cancer (GIRG) guidelines for gastric cancer staging and treatment: 2015. Gastric Cancer. 2017;20:20–30.
Challine A, Voron T, Dousset B, et al. Postoperative outcomes after laparoscopic or open gastrectomy: A national cohort study of 10,343 patients. Eur J Surg Oncol. 2021;47:1985–1995.
Baiocchi GL, Giacopuzzi S, Reim D, et al. Incidence and grading of complications after gastrectomy for cancer using the GASTRODATA Registry. Ann Surg. 2020;272:807–813.
Voeten DM, Busweiler LAD, van der Werf LR, et al. Outcomes of esophagogastric cancer surgery during eight years of surgical auditing by the Dutch Upper Gastrointestinal Cancer Audit (DUCA). Ann Surg. 2021;274:866–873.
Papenfuss WA, Kukar M, Oxenberg J, et al. Morbidity and mortality associated with gastrectomy for gastric cancer. Ann Surg Oncol. 2014;21:3008–3014.
Watanabe N, Miyata H, Gotoh M, et al. Total gastrectomy risk model: data from 20,011 Japanese patients in a nationwide internet-based database. Ann Surg. 2014;260:1034–1039.
Sallehi O, Vega EA, Kutlu OC, et al. Western population-based study of oncologic surgical quality and outcomes of laparoscopic versus open gastrectomy for gastric adenocarcinoma. Surg Endosc. 2021;35:4786–4793.
D´Journo XB, Boulate D, Fourdrain A, et al. Risk prediction model of 90-day mortality after esophagectomy for cancer. JAMA Surg. 2021;156:836–845.
Fischer C, Lingsma H, Hardwick R, et al. Risk adjustment models for short-term outcomes after surgical resection for oesophagogastric cancer. Br J Surg. 2016;103:105–106.
Van den Boorn HG, Engelhardt EG, van Kleef J, et al. Prediction models for patients with esophageal or gastric cancer: a systematic review and meta-analysis. PLoS ONE. 2018;13:e0192310.
Kurita N, Miyata H, Gotoh M, et al. Risk model for distal gastrectomy when treating gastric cancer on the basis of data from 33,917 Japanese patients collected using a nationwide web-based data entry system. Ann Surg. 2015;262:295–303.
Collins GS, Dhiman P, Andaur Navarro C, et al. Protocol for development of a reporting guideline (TRIPOD-AI) and risk of bias tool (PROBAST-AI) for diagnostic and prognostic prediction model studies based on artificial intelligence. BMJ Open. 2021;11:e048008.
Dhiman P, Ma J, Andaur-Navarro C, et al. Reporting of prognostic clinical prediction models based on machine learning methods in oncology needs to be improved. J Clin Epidemiol. 2021;138:60–72.
Andaur-Navarro CL, Damen JA, Takada T, et al. Risk of bias in studies on prediction models developed using supervised machine learning techniques: a systematic review. BMJ. 2021;375:n281.
Dal Cero M, Rodríguez-Santiago J, Miro M, et al. Evaluation of data quality in the Spanish EURECCA Esophagogastric Cancer Registry. Eur J Surg Oncol. 2021;47:3081–3087.
Collins GS, Reitsma JB, Altman DG, et al. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. Ann Intern Med. 2015;162:55–63.
Baiocchi GL, Giacopuzzi S, Marrelli D, et al. International consensus on a complications list after gastrectomy for cancer. Gastric Cancer. 2019;22:172–189.
Dindo D, Demartines N, Clavien PA. Classification of surgical complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann Surg. 2004;240:205–213.
Hancock JT, Khoshgoftaar TM. Survey on categorical data for neural networks. J Big Data. 2020;7:28.
Ding Y, Simonoff JS. An investigation of missing data methods for classification trees applied to binary response data. J Mach Learn Res. 2010;11:131–170.
Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models vi coordinate descent. J Stat Softw. 2010;33:1–2.
Breiman L. Random forest. Mach Learn. 2001;45:5–32.
Bühlmann P, Yu B. Boosting with the L2 loss: regression and classification. J Am Statist Assoc. 2003;98:324–339.
Schapire RE. The strength of weak learnability. Mach Learn. 1990;5:197–227.
Chawla NV, Bowyer KW, Hall LO, et al. SMOTE: synthetic minority over-sampling technique. J Artif Intell Res. 2002;6:321–357.
Cawley GC, Talbot NLC. On over-fitting in model selection and subsequent selection bias in performance evaluation. J Mach Learn Res. 2010;11:2079–2107.
Chen W, Sahiner B, Samuelson F, et al. Calibration of medical diagnostic classifier scores to the probability of disease. Stat Methods Med Res. 2018;27:1394–1409.
Biecek P. DALEX: Explainers for complex predictive models in R. J Mach Learn Res. 2018;19:1–5.
Lang M, Binder M, Richter J, et al. mlr3: A modern object-oriented machine learning framework in R. J Open Source Softw. 2019;4:1903.
Shannon AB, Straker RJ, Fraker DL, et al. Ninety-day mortality after total gastrectomy for gastric cancer. Surgery. 2021;170:603–609.
Hamilton TD, Mahar A, Haas B, et al. The impact of advanced age on short-term outcomes following gastric cancer resection: an ACS-NSQIP analysis. Gastric Cancer. 2018;21:710–719.
Iswaran H, O´Brien R. Commentary: the problem of class imbalance in biomedical data. J Thorac Cardiovasc Surg. 2021;161:1940–1941.
Khushi M, Shaukat K, Alam TM, et al. A comparative performance analysis of data resampling methods on imbalance medical data. IEEE Access. 2021;9:109960–109975.