Skip to main content

Important factors for the modelling and design of clinical trials for severe sepsis and multiple organ failure

Introduction

No large, well-controlled, trial has been able to demonstrate a statistically significant and reproducible benefit of experimental treatment in severe sepsis and multiple organ failure. This study was done to determine the factors that have to be controlled for in future design of clinical trials in sepsis.

Method

2790 patients from the RIP database satisfied the criteria of severe sepsis and multiple organ failure. Logistic regression analysis was carried out to determine the factors that influenced ICU outcome. The CHAID model of an expert system AnswerTree (SPSS, UK) was also used to derive decision rules that govern the outcome of these patients.

Results

Of the eight independent variables entered into the logistic regression analysis four in order of importance were selected: APACHE II score on the day of development of sepsis, treating centre, number of organ failures, age. The area under ROC was 0.75. The level and branches of the decision rules by the expert system is shown in the Figure on the previous page. The difference in outcome for all the nodes is P < 0.0001.

figure 1

Conclusion

As the area under the curve of the ROC = 0.75, one is unlikely to use logistical regression analysis to risk stratify patients for future trials of severe sepsis; however, expert systems can delineate statistical significance and patterns which influence outcome in a complex trial population.

Author information

Authors and Affiliations

Authors

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Shaikh, L., Stuart, H., Rhodes, A. et al. Important factors for the modelling and design of clinical trials for severe sepsis and multiple organ failure. Crit Care 3 (Suppl 1), P255 (2000). https://doi.org/10.1186/cc628

Download citation

  • Published:

  • DOI: https://doi.org/10.1186/cc628

Keywords

  • Clinical Trial
  • Logistic Regression
  • Logistic Regression Analysis
  • Emergency Medicine
  • Severe Sepsis