American Journal of Epidemiology Advance Access originally published online on August 17, 2006
American Journal of Epidemiology 2006 164(7):707-708; doi:10.1093/aje/kwj292
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Letter to the Editor |
RE: "THE INCONSISTENCY OF ‘OPTIMAL’ CUTPOINTS OBTAINED USING TWO CRITERIA BASED ON THE RECEIVER OPERATING CHARACTERISTIC CURVE"
Department of Clinical Chemistry and Transfusion Medicine, Sahlgrenska University Hospital, S-413 45 Göteborg, Sweden
(e-mail: henrik.zetterberg{at}clinchem.gu.se)
Perkins and Schisterman (1
) are to be congratulated on an important contribution toward advancing the accuracy of optimal biomarker cutpoint selection in clinical diagnosis of disease. They compared two methods that are commonly used for this purpose, the closest-to-(0,1) criterion in the receiver operating characteristic curve and the Youden index, J. They concluded that the Youden index is to be preferred and cautioned against the use of the closest-to-(0,1) criterion, since the latter introduces an increased rate of misclassification, exemplified in a data set on preeclampsia.
I would like to raise one additional point of caution: The literature contains different definitions of the Youden index. According to most studies, the Youden index is defined as sensitivity plus specificity minus 1, but other authors state that it is the positive predictive value plus the negative predictive value minus 1 (2). When examining Youden's original publication (3), it is clear, after some calculations, that the index indeed is sensitivity plus specificity minus 1. This is also the definition used by Perkins and Schisterman (1).
The Youden index is also used for rating diagnostic tests. However, for this purpose the index adds little to sensitivity and specificity figures alone. On the other hand, a Youden index based on predictive values would be more informative, since it directly incorporates the powerful influence of disease prevalence on the usefulness of a diagnostic test. A hypothetical evaluation of two qualitative diagnostic tests (A and B) shows that the different ways of calculating the index indeed make a difference (table 1). The Youden index indicates that test B is superior to test A, while the index based on predictive values indicates the opposite.
|
From a clinical perspective, the latter view may be considered the most correct. I propose that a predictive-value-based Youden index be considered in addition to J when evaluating cutpoints and rating diagnostic tests in clinically relevant populations.
ACKNOWLEDGMENTS
Conflict of interest: none declared.
References
- Perkins NJ, Schisterman EF. The inconsistency of "optimal" cutpoints obtained using two criteria based on the receiver operating characteristic curve. Am J Epidemiol 2006;163:670–5. (Epub 2006 Jan 12).
[Abstract/Free Full Text] - Grimes DA, Schulz KF. Uses and abuses of screening tests. Lancet 2002;359:881–4.[CrossRef][ISI][Medline]
- Youden WJ. Index for rating diagnostic tests. Cancer 1950;3:32–5.[CrossRef][ISI][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  N. J. Perkins and E. F. Schisterman THE AUTHORS REPLY Am. J. Epidemiol., October 1, 2006; 164(7): 708 - 708. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||