Letters Section Editor: Stephen J. Lurie,
MD, PhD, Senior Editor.
To the Editor: Although the study by Dr Greenland
and colleagues1 documented the high frequency
of at least 1 of the 4 major risk factors among patients with CHD, their data
provide no clear measure of the power of this risk profile to predict high
risk of CHD events.
Using data from the authors' Table 1, I used likelihood ratio (LR) analysis2- 5 to
compute the predictive power of having at least 1 risk factor. The sensitivity,
specificity, and complement of specificity (100 − specificity) were
derived for the combined data of the 3 studies that Greenland et al described.
The likelihood ratio for an event (ie, a positive LR) in odds:1 units is a
prevalence-independent measure calculated as the sensitivity divided by (100
− specificity). An LR of 2.0 or less denotes low predictive power and
an LR of 9.0 or more denotes high predictive power. Table 1 shows the composite measures for the 3 studies of the predictive
power for CHD death, or for CHD death or nonfatal myocardial infarction, in
men and women ages 18 to 59 years, as determined by the presence of 1 or more
Weissler AM. Traditional Risk Factors for Coronary Heart Disease. JAMA. 2004;291(3):299. doi:10.1001/jama.291.3.299-c