Clinical predictors of ongoing infection in secondary peritonitis: systematic review

Abstract

Introduction: The decision to perform a relaparotomy in patients with secondary peritonitis is based on "clinical judgment" with inherent variability among surgeons. Our objective was to review the literature on prognostic variables for ongoing abdominal infection. Predictive variables for positive findings at relaparotomy can generate more objective criteria to support the decision whether to perform a relaparotomy in patients with secondary peritonitis.

Methods: Multiple databases were searched for studies assessing the prognostic value of clinical variables predicting outcome of relaparotomy or general outcome in patients with secondary peritonitis. Data on the methodologic quality of the study as well as statistical strength of predictors and validity of individual variables were extracted and scored. A cumulative score was calculated from these three scores, and the variables were ranked.

Results: A total of 37 of 197 retrieved articles were included for final assessment. The median score for methodologic quality of individual articles was 36 (range 19-54). After calculation of the combined scores, 76 individual variables (patient, peritonitis, surgery, clinical, and laboratory variables) were identified from which the top 10 were eventually selected. These variables were age, concomitant disease, upper gastrointestinal source of peritonitis, generalized peritonitis, elimination of the focus, bilirubin, creatinine, lactate, PaO2/FiO2 ratio, and albumin. This set of variables proved to be moderately predictive for positive findings during relaparotomy in a retrospective cohort of 219 patients operated on for secondary peritonitis (receiver operator curve 0.75, with 95% confidence interval 0.68-0.82).

Conclusions: This review generated a hierarchy (weighted ranking) of published variables that could play a role in the decision to perform a relaparotomy in patients with secondary peritonitis. The top sixtile of ranked variables (10 variables) showed promising results in the discrimination between patients having a positive and negative relaparotomy when tested on a peritonitis patient database. This ranking of variables provides evidence for potential inclusion of variables in future predictive scores, although improvement in overall predictive strength of a set of variables in such a score is needed.