Accuracy of risk assessment in primary prevention of cardiovascular disease.

Klaus Eichler, Research fellow,

Other Contributors:

June 08, 2006

Dear Editor,

The systematic review by Brindle et al. [1] about the accuracy and impact of the Framingham risk score in the primary prevention of cardiovascular disease (CVD) is an important contribution. It summarizes the evidence about the predictive performance of this score in different populations but has some limitations. We have four observations.

First, the applied inclusion criteria (for review part A) were not consistent with the primary prevention setting in clinical practice. Some cohorts included up to 20% of persons with established CVD qualifying for secondary prevention or active intervention groups of clinical trials, which are treated according to a standardised protocol in contrast to primary care populations. It is problematic to evaluate calibration in cohorts not being representative for the population of interest.

Second, the authors excluded studies that reported only fatal outcomes even though Framingham functions for coronary death exist. [2] It can be argued that current guidelines using Framingham derived scores refer to hard coronary events (myocardial infarction and coronary death) rather than coronary death alone. However, to summarize the overall evidence also studies reporting only data for coronary death seem appropriate for analysis because these data are highly correlated to hard coronary events. [2]

Third, the authors did not apply a formal quality assessment of the included studies that evaluated calibration. But recommendations for methodological standards exist (e.g. blinding of outcome assessment; completeness and sufficient length of follow-up; or description of population variables). [3, 4]

Fourth, the authors do not report a formal assessment of heterogeneity of calibration data and conclude that the accuracy of the Framingham score relates to the background risk of the validation cohorts (i.e. under-prediction of risk in higher risk cohorts and vice versa). However, little information is given about additional factors that may confound the influence of background risk on performance of the risk score.

In conclusion, the systematic review by Brindle et al. [1] is an important contribution showing the variable calibration of the Framingham score across different populations but also lacks a rigorous evaluation of factors which may have influence on calibration. We believe that further work is needed to learn more about factors influencing predictive performance of the Framingham risk score across different populations, which can have direct implications for use in clinical practice.

Competing interests All authors declare that the answer to the questions on your competing interest form are all No and therefore have nothing to declare.

Copyright “The Corresponding Author has the right to grant on behalf of all authors and does grant on behalf of all authors, an exclusive licence (or non exclusive for government employees) on a worldwide basis to the BMJ Publishing Group Ltd and its licensees, to permit this article (if accepted) to be published in Heart and any other BMJPG products and to exploit all subsidiary rights, as set out in our licence (”

Klaus Eichler1
Milo A Puhan1
Johann Steurer1
Lucas M. Bachmann1

1 Horten Centre for patient oriented research,
University hospital of Zurich,

Corresponding author:

Klaus Eichler, MD MPH
Horten Centre
Zurich University
Postfach Nord
Tel.: +41 1 255 31 06 (86 62)
Fax.: +41 1 255 97 20


1. Brindle PM, Beswick AD, Fahey TF, Ebrahim SB: The accuracy and impact of risk assessment in the primary prevention of cardiovascular disease: A sytematic review. Heart 2006.

2. Anderson KM, Odell PM, Wilson PW, Kannel WB: Cardiovascular disease risk profiles. Am Heart J 1990, 121:293-8.

3. Laupacis A, Sekar N, Stiell IG: Clinical prediction rules. A review and suggested modifications of methodological standards. JAMA 1997, 277:488- 94.

4. Altman DG: Systematic reviews of evaluations of prognostic variables. British Medical Journal 2001, 323:224-8.

Conflict of Interest

None declared