Premium
Preoperative assessment of mortality risk in hepatic resection by clinical variables: A multivariate analysis
Author(s) -
Bolder Ulrich,
Brune Andreas,
Schmidt Silke,
Tacke Jürgen,
Jauch KarlWalter,
Löhlein Dietrich
Publication year - 1999
Publication title -
liver transplantation and surgery
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.814
H-Index - 150
eISSN - 1527-6473
pISSN - 1074-3022
DOI - 10.1002/lt.500050302
Subject(s) - medicine , multivariate analysis , logistic regression , multivariate statistics , cirrhosis , receiver operating characteristic , surgery , risk factor , mathematics , statistics
Hepatic resection is a chance for cure for primary and secondary liver tumors and a variety of benign diseases. Despite advances in surgical technique and patient care, preoperative and postoperative morbidity in patients undergoing liver resection remains high. Because a high morbidity represents a risk factor contributing to a fatal outcome of the surgical procedure, our study aimed to investigate the contribution of different risk factors to a fatal outcome and if mortality can be predicted by the presence of certain risk factors. Two hundred fifty‐seven patients undergoing hepatic resection (curative and palliative) were analyzed preoperatively, immediately after surgery, and 10 days after surgery for 60 potential risk factors. Survivors (n = 238) and nonsurvivors (n = 19) were compared univariately. The analysis identified 14 variables to differentiate between groups. These variables were processed by multivariate logistic regression analysis. Three models to estimate 30‐day mortality were identified, tested for statistical accuracy, and assessed for their receiver‐operated characteristics (ROCs). The variables in the multivariate models were as follows: preoperatively, age, number of comorbid factors, and presence of cirrhosis; immediately after surgery, age, number of comorbid factors, and percentage of resected liver; and 10 days after surgery, age, hours of ventilation, and number of adverse events. Goodness of fit was 0.863, 0.912, and 0.966, respectively. Areas under the ROC curves were 83.6%, 85.7%, and 98.0%. The specificity (probability to identify survivors correctly) was greater than 90% for all models, although sensitivity (probability to identify nonsurvivors correctly) was greater than 90% only for 10 days after surgery. We conclude that logistic regression is appropriate to assess the importance of risk factors in the course of hepatic resection and to identify patient groups at high risk.