Author Affiliations: Mount Sinai School of Medicine, New York, New York (Dr Domanski) (email@example.com); and Duke Clinical Research Institute, Durham, North Carolina (Drs Mahaffey and Hasselblad).
In Reply: Dr Conti and colleagues are concerned that the relationship between CK-MB levels and 30-day mortality was not linear. We explored a wide range of models that would relate CK-MB levels to mortality. Any model chosen had to easily be explained and be able to be compared with other models (with and without covariates) across different time periods. Splines, which are parametric, have a different function for different ranges of the independent variable. For example, a 3-knot spine has 4 different functions over 4 intervals. Polynomials can be unstable, especially for small and large independent variable values, and are difficult to compare across analyses. More sophisticated models, such as the multistage model and gamma multihit model, require several terms and are difficult to compare across analyses.
Domanski MJ, Mahaffey K, Hasselblad V. Myocardial Enzyme Levels and Mortality After Coronary Artery Bypass Graft Surgery—Reply. JAMA. 2011;306(1):39-40. doi:10.1001/jama.2011.896