Ann Thorac Surg 2006;81:2088
© 2006 The Society of Thoracic Surgeons
Original article: Cardiovascular
Invited commentary
Luis D. Berrizbeitia, MD
Department of Surgery (Thoracic), University Medical Center at Princeton, 253 Witherspoon St, Suite F, Princeton, NJ 08540
(Email: ldberriz{at}ctsx.net).
Accurate estimation of mortality risk is an essential component of complex cardiac operations such as replacement of the ascending aorta. A number of risk stratification models have been developed to assist the clinician in estimating the risk of surgery [1]. Under ideal circumstances, given a number of known factors, a risk stratification model will predict whether a particular patient survives or dies after an operation. Unfortunately, in real-life situations, risk models explain only a small proportion of the variability in outcomes and characteristically perform less well when applied to populations different from the one from which they were developed [2].
Matsuura and associates [3] infer by observation of their own clinical results that the EuroSCORE predicts higher than actual mortality in patients undergoing ascending aortic repair and that removal of the variable age improves the accuracy of the model. In carrying out this analysis, the authors have engaged in an exercise of model validation by calibration and discrimination. Calibration refers to the ability for a model to predict mortality accurately. Discrimination refers to the ability to distinguish patients who die from those who survive [4]. By removing the variable age they demonstrate improved calibration by closer agreement between the observed and predicted mortality by risk categories, and improved discriminatory power by a greater area under the curve of the receiver operator characteristic curve.
Cardiac surgeons may be tempted by these findings to disregard age as a risk factor in ascending aortic surgery for which it is important to consider some of the limitations of the study. First, the data presented for model calibration lacks statistical analysis and the results depend on a particular selection of risk categories. This data should be considered a valuable observation but not an actual finding. Second, the underlying methodology to construct receiver operator characteristic curves and determine the area under the curve is nonparametric. It relies on the ranks of a set of observations to present the average discriminatory power of the model. This means that in the absence of a formal and comprehensive statistical analysis of the model calibration, the area under the curve alone gives a limited representation of the validity of the revised model [4].
Regardless of the specifics of the analysis, this study demonstrates that surgery for ascending aortic repair has evolved considerably during the last several decades and that it can be safely applied to a greater proportion of elderly patients.
 |
References
|
|---|
- Ferraris VA, Ferraris SP. Risk stratification and comorbidityIn: Cohn LH, Edmunds LH, editors. Cardiac surgery in the adult, 2nd ed. New York: McGraw-Hill; 2003. pp. 187-224.
- Shahian DM, Blackstone EH, Edwards FH, et al. Cardiac surgery risk models: a position article Ann Thorac Surg 2004;78:1868-1877.[Abstract/Free Full Text]
- Matsuura K, Ogino H, Matsuda H, et al. Limitations of EuroSCORE for measurement of risk-stratified mortality in aortic arch surgery using selective cerebral perfusionis advanced age no longer a risk?. Ann Thorac Surg 2006;81:2084-2088.[Abstract/Free Full Text]
- Omar RZ, Ambler G, Royston P, Eliahoo J, Taylor KM. Cardiac surgery risk modeling for mortality: a review of current practice and suggestions for improvement Ann Thorac Surg 2004;77:2232-2237.[Abstract/Free Full Text]
Top
References