Skip Navigation

Electronic Letters to:

Heart failure/cardiomyopathy:
Rafael Vazquez, Antoni Bayes-Genis, Iwona Cygankiewicz, Domingo Pascual-Figal, Lilian Grigorian-Shamagian, Ricardo Pavon, Jose R. Gonzalez-Juanatey, José M. Cubero, Luis Pastor, Jordi Ordonez-Llanos, Juan Cinca, and Antoni Bayes de Luna on behalf of the MUSIC Investigators
The MUSIC Risk score: a simple method for predicting mortality in ambulatory patients with chronic heart failure
Eur Heart J 2009; 30: 1088-1096 [Abstract] [Full text] [PDF]
*E-letters: Submit a response to this article

Electronic letters published:

[Read eLetter] The MUSIC Risk score: the relevance of competing risks analysis
Darryl P. Leong, Amy B. Salter, and Christian A. Gericke   (3 November 2009)

The MUSIC Risk score: the relevance of competing risks analysis 3 November 2009
Top
Darryl P. Leong,
Clinical Lecturer
University of Adelaide, Adelaide, Australia 5000,
Amy B. Salter, and Christian A. Gericke

Send e-letter to journal:
Re: The MUSIC Risk score: the relevance of competing risks analysis

We wish to comment on the study by Vazquez et al. that models mortality risk in ambulatory chronic heart failure patients[1]. We raise profound concerns over the statistical methodology employed. Cox proportional hazards modelling was used to determine the contribution of covariates to the risk of all-cause mortality (n=267), cardiac mortality (primary end-point, n=213) and two subgroups of cardiac mortality, pump-failure death (n=123) and sudden death (n=90). In each case this technique ignores the competing risk of non-cardiac death. In the study sample, approximately one quarter of the cohort that died did so from non-cardiac causes. Cox regression with a time to cardiac death outcome makes the incorrect assumption that all individuals included in the model can only die from cardiac causes. In this study individuals who died for other reasons or who did not die during the study period were both treated as censored observations. This means that the proposed MUSIC risk score does not predict mortality in chronic heart failure patients as suggested by the authors but merely represents a hypothetical construct which does not represent real-life circumstances and therefore lacks both internal and external validity. Cox regression may be acceptable in studies in which the statistical model narrowly reflects a biological mechanism or when only a single event of interest occurs[2], but is inappropriate in the development of a clinical tool to predict risk of death when competing risks exist. In scenarios where endpoints include competing risks – in this instance cardiac and non- cardiac death - a competing risk model, such as that proposed by Fine and Gray[3] provides more appropriate estimation of risk of the outcome of interest. Ignoring competing risks can result in grossly inaccurate estimation of probabilities[4]. The error that can arise in failing to take competing risks into account is readily demonstrable in simple hypothetical scenarios.

Lastly, there appears to be incomplete information about loss-to- follow-up. The authors state that 11 subjects were lost to follow-up and therefore censored, but no further mention is made of these subjects in the results section.

References

1. Vazquez, R., et al., The MUSIC Risk score: a simple method for predicting mortality in ambulatory patients with chronic heart failure. Eur Heart J, 2009. 30(9): p. 1088-96.

2. Pintilie, M., Competing Risks: A Practical Perspective. 2006, Wiley: Chichester. p. 53-55.

3. Fine, J.P. and R.J. Gray, A proportional hazards model for the subdistribution of a competing risk. Journal of the American Statistical Association, 1999. 94: p. 496-509.

4. Kalbfleisch, J.D. and R.L. Prentice, The Statistical Analysis of Failure Time Data. 1980, New York: Wiley.

Conflict of Interest:

None declared