Dynamic prognostication in non-ST–elevation acute coronary syndromes: insights from GUSTO-IIB and pursuit

Chang, Wei-Ching; Boersma, Eric; Granger, Christopher B; Harrington, Robert A; Califf, Robert M; Simoons, Maarten L; Kleiman, Neal S; Armstrong, Paul W; GUSTO-IIb and PURSUIT Investigators,

doi:10.1016/j.ahj.2003.05.004

« Previous Next »American Heart Journal
Volume 148, Issue 1 , Pages 62-71, July 2004

Dynamic prognostication in non-ST–elevation acute coronary syndromes: insights from GUSTO-IIB and pursuit☆

Wei-Ching Chang, PhD
- University of Alberta, Edmonton, Alberta, Canada
Eric Boersma, PhD
- Erasmus University, Rotterdam, The Netherlands
Christopher B Granger, MD
- Duke Clinical Research Institute, Durham, NC, USA
Robert A Harrington, MD
- Duke Clinical Research Institute, Durham, NC, USA
Robert M Califf, MD
- Duke Clinical Research Institute, Durham, NC, USA
Maarten L Simoons, MD
- Erasmus University, Rotterdam, The Netherlands
Neal S Kleiman, MD
- Baylor College of Medicine, Houston, Tex, USA
Paul W Armstrong, MD
- University of Alberta, Edmonton, Alberta, Canada
- Reprint requests: Paul W. Armstrong, MD, Department of Medicine, 2-51 Medical Sciences Building, University of Alberta, Edmonton, Alberta T6G 2H7 Canada.
GUSTO-IIb and PURSUIT Investigators

Received 27 December 2002; accepted 14 May 2003.

Abstract

Background

Risk assessment in patients with non-ST–elevation acute coronary syndromes (NSTE-ACS) traditionally focuses on and is limited to admission findings. The objective of the current study was to develop an approach to predicting outcome in NSTE-ACS that could account for the changing nature of risk.

Methods

In 7294 of 8010 patients with NSTE-ACS and complete electrocardiographic data in the GUSTO-IIb trial, we predicted the mortality probability at days 0–2, 0–30, 3–30, 5–30, and 7–30 using multiple logistic regression. Resulting risk estimates were incorporated into a composite, dynamic model to estimate the effects of changing probabilities over time. These models were validated against an independent sample of 9461 patients from the PURSUIT trial.

Results

As time passed after admission, the risk of 30-day death declined in stable patients. This risk, which was 3.72% at baseline, declined to 1.92% in 6-day survivors, and the risk reduction was greatest for those with the highest baseline risk. Importantly, however, the development of inhospital complications modified these trends. The use of dynamic models not only allowed us to estimate early (<48 h) mortality with a high degree of accuracy (C-index of 0.87), but also to continuously update the longer-term prognosis with increasing accuracy: the C-index increased from 0.75 for the day 0–30 model to 0.81 and 0.82 for the composite and day 7–30 models, respectively.

Conclusions

Dynamic risk assessment is feasible and reliable. This approach can improve risk assessment and provide valuable guidance for management of patients with NSTE-ACS.