Predicting Survival in Heart Failure: A Risk Score Based on 39,372 Patients From 30 Studies

Nov 06, 2012
Font Size
A
A
A
Authors:
Pocock SJ, Ariti CA, McMurray JJ, et al., on behalf of the Meta-Analysis Global Group in Chronic Heart Failure (MAGGIC).
Citation:
Eur Heart J 2012;Oct 24:[Epub ahead of print].
Summary By:
Jennifer Ann Cowger, MD, MS, FACC

Study Questions:

Can a risk score for predicting heart failure mortality be devised?

Methods:

MAGGIC (Meta-Analysis Global Group in Chronic Heart Failure) was a meta-analysis of 39,372 patients with a diagnosis of heart failure included in 30 cohort studies (six randomized controlled trials). Poisson regression was used to identify predictors of death within 3 years, and a risk model was developed.

Results:

Mortality was 40.2% (n = 15,851) over a median 2.5 years of follow-up. Predictors of mortality included older age, male sex, smaller body mass index, tobacco use, lower systolic blood pressure, higher New York Heart Association class, lower left ventricular ejection fraction (LVEF), longer heart failure duration, diabetes, chronic obstructive pulmonary disease, higher creatinine, absence of beta-blocker use, and absence of angiotensin-converting enzyme inhibitor use. In patients with higher ejection fractions, the impact of age was greater and blood pressure was less. The MAGGIC score ranges from 0-52, and the cohort median value as 23. The 3-year probabilities of death in patients with scores of 10, 20, 30, and 40 were 0.10, 0.26, 0.53, and 0.84, respectively. Predicted and observed mortalities were similar.

Conclusions:

A mortality risk prediction score was devised for patients with systolic and diastolic heart failure.

Perspective:

This analysis used trial data to develop a risk prediction model for patients with heart failure of either systolic or diastolic etiology. For some significant predictors, patient level data were missing in a large percentage (e.g., 20,045 of 39,372 were missing creatinine values). As the authors pointed out, the impact of many variables can be different based on heart failure etiology (i.e., preserved vs. reduced LVEF) and, therefore, it is not clear that ‘lumping’ all heart failure prognostication into one model is the proper technique. Model validation is also required.

Keywords: Risk, Follow-Up Studies, Blood Pressure, New York, Heart Diseases, Body Mass Index, Cardiology, Heart Failure, Stroke Volume, Tobacco Use, Diabetes Mellitus, Lung Diseases


< Back to Listings