[Analysis of risk factors for gastrointestinal bleeding in patients undergoing extracorporeal membrane oxygenation and establishment of a predictive model].

OBJECTIVE

To investigate the risk factors for gastrointestinal bleeding during extracorporeal membrane oxygenation (ECMO) support, construct a risk prediction nomogram model, and validate its clinical application efficacy.

METHODS

Critically ill patients who received ECMO therapy in the department of critical care medicine of the First Affiliated Hospital of Wannan Medical College from December 2016 and December 2024 were retrospectively enrolled. Univariate analysis was employed to examine the influencing factors related to gastrointestinal bleeding in patients. Multivariate Logistic regression analysis was used to identify independent risk factors for gastrointestinal bleeding during ECMO. Based on these factors, a risk prediction nomogram model was constructed. The goodness-of-fit and clinical performance of the model was evaluated by receiver operator characteristic curve (ROC curve), calibration curve, and decision curve analysis (DCA) curve, respectively.

RESULTS

A total of 189 ECMO patients were included, with an average age of (53.65±15.66) years (12-75 years old); gastrointestinal bleeding occurred in 46 patients (24.3%), including 12 cases (6.3%) on veno-venous (V-V) ECMO, 33 cases (17.5%) on veno-arterial (V-A) ECMO, and 1 case (0.5%) on veno-venous-arterial (V-V-A) ECMO. Successful ECMO weaning was achieved in 109 patients, while weaning failed in 80 patients. Outcomes included improvement in 79 patients, no improvement in 39 patients, and death in 71 patients. The factors significantly associated with gastrointestinal bleeding in ECMO patients in univariate analysis [septic shock, alcohol consumption history, extracorporeal cardiopulmonary resuscitation (ECPR), platelet count (PLT), and activated partial thromboplastin time (APTT)] were included in the multivariate Logistic regression analysis, showed that septic shock [odds ratio (OR)=33.590, 95% confidence interval (95%CI) was 5.640-200.055, P<0.001], alcohol consumption history (OR=9.016, 95%CI was 2.251-36.119, P=0.002), ECPR (OR=49.440, 95%CI was 2.572-950.292, P=0.010), PLT decrease (OR=0.981, 95%CI was 0.966-0.996, P=0.013), and APTT extension (OR=1.041, 95%CI was 1.019-1.062, P<0.001) were independent risk factors for gastrointestinal bleeding. A nomogram model was constructed using these five independent risk factors, the ROC curve showed that the nomogram model predicted gastrointestinal bleeding with an area under the curve of 0.96 (95%CI was 0.93-1.00), sensitivity of 94%, and specificity of 88%. The calibration curve demonstrated good consistency between the nomogram-predicted probability of gastrointestinal bleeding and actual outcomes. The DCA indicated that the nomogram provided clinical net benefit across a wide range of threshold probabilities, confirming its practical utility.

CONCLUSIONS

Careful management during ECMO is crucial. The prediction model incorporating septic shock, alcohol consumption history, ECPR, PLT decrease, and APTT extension can be constructed to screen high-risk patients for gastrointestinal bleeding, in order to improve the success rate of ECMO treatment.