Multivariate analysis of 9 disease-associated variables for outcome prediction in patients with sepsis
S E Calvano1, S M Coyle, K S Barbosa
1Department of Surgery, University of Medicine and Dentistry of New Jersey-Robert Wood Johnson Medical School, New Brunswick 08903, USA. calvanst@umdnj.edu
Objective:
To assess the ability of 9 clinical or biological variables to predict outcome (survival or nonsurvival) using multiple regression and classification analyses.
Design:
Prospective, descriptive cohort study with no interventions.
Setting:
Surgical intensive care unit of a tertiary care hospital and a medical school research laboratory.
Patients:
Eighteen patients with a documented source of infection who met currently accepted criteria for sepsis syndrome or septic shock.
Main Outcome Measures:
Prediction of survival or nonsurvival based on analysis of clinical (Multiple Organ Dysfunction score, Acute Physiology and Chronic Health Evaluation III scores) and biological (plasma levels of cortisol, interleukin 6, interleukin 10, phospholipase A2, soluble tumor necrosis factor receptor p75, and monocyte membrane tumor necrosis factor receptor levels) variables, with comparison of predicted and actual outcomes.
Results:
Plasma interleukin 6, interleukin 10, and phospholipase A2 concentrations were not significantly (P>.05) different between survivors and nonsurvivors. By standard, forward stepwise, and backward stepwise multiple regression analyses, only monocyte membrane tumor necrosis factor receptor levels measured at the onset of sepsis significantly predicted outcome in all 3 analyses. However, by both standard and backward stepwise analyses, Multiple Organ Dysfunction scores based on evaluation at the onset of sepsis and 24 hours later were also significant predictors of outcome. Classification analysis showed that assignment to outcome group was statistically significant when based on monocyte membrane tumor necrosis factor receptor levels determined at the onset of sepsis or on Multiple Organ Dysfunction scores assessed 24 hours after sepsis was diagnosed.
Conclusion:
Although these findings were based on a relatively small cohort, both multiple regression and classification analyses indicated that only monocyte membrane tumor necrosis factor receptor levels are able to discriminate survivors from nonsurvivors at the onset of sepsis.
