Affiliations: Department of Public Health and Primary Care and
Department of Biostatistics, University of Cambridge, Cambridge, UK | Northwestern University Clinical and Translational
Sciences Institute and Department of Preventive Medicine at Northwestern
University Feinberg School of Medicine, Chicago, IL, USA
Note: [] Corresponding author: Dr. Greenland, 750 North Lake Shore Drive,
11th Floor, Chicago, IL 60611, USA
Abstract: New predictors of cardiovascular outcomes are widely sought in
research settings, and predictive tests are commonly recommended for routine
use in cardiovascular clinical care. A number of multivariable scoring systems
are in use around the world for assessment of a patient's risk. While such
scoring systems are often recommended for clinical use in medical practice
guidelines, their actual use in medical care falls short of recommendations.
Limitations in the predictive capacity of existing predictive models are
recognized, including lack of predictive accuracy, lack of ability to separate
those who develop events from those who do not, and risks and costs of the
testing modalities. Biomarker research is actively developing new testing
strategies trying to improve upon current approaches, but it is often unclear
how to assess the incremental prognostic information that a new test provides.
In this report, we discuss the statistical approaches that can be used to
evaluate additive predictive value of new tests. We also consider clinical
research examples to put this information into a practical context.
