July 11, 2007

Algorithms for Assessing Cardiovascular Risk in Women

JAMA. 2007;298(2):175-178. doi:10.1001/jama.298.2.176-b

To the Editor: Dr Ridker and colleagues1 report good results for their cardiovascular risk-scoring method for women, the Reynolds Risk Score. However, the comparison they make to a risk score published by the Framingham Study group is unfairly weighted against the Framingham score. The “validation data” on which the 2 scores are evaluated are closely related to the data on which the Reynolds score is built, having been recruited simultaneously from the same population with the same inclusion criteria and recruitment procedures, and separated from the “training data” only by random resampling. This data set therefore has a close relation to the Reynolds score that it does not have with the Framingham score. The Reynolds score is being subjected to a “cross-validation” procedure while the Framingham score is subjected to “external validation”, regarded as a much sterner test.2

First Page Preview View Large
First page PDF preview
First page PDF preview