Valid XHTML     Valid CSS2    

A simple imperfect gold standard computation

                     gilles.hunault "at" univ-angers.fr

Enter the sensitivity of the marker relative to true disease, for instance 0.8

            

Enter the value (1-specificity) of the marker relative to true disease, for instance 0.7

            

Enter the sensitivity of the biopsy relative to true disease, for instance 0.95

            

Enter the specificity of the biopsy relative to true disease, for instance 0.9

            

Enter the prevalence of true disease, for instance 0.4

            

                    clear all values               use the reference values 

Reference

Exceeding the limits of liver histology markers

     Shruti H. Mehta, Bryan Lau, Nezam H. Afdhal, David L. Thomas

     Journal of Hepatology, Volume 50, Issue 1, January 2009, Pages 36-41.

Programmed formula

     
non su
     

D+ and D denote the true stage of disease; B+ and B indicate biopsy diagnosed disease; M denotes the value on a continuous scale of a marker; μB:D and γB:D are the sensitivity and specificity of the biopsy relative to true disease, respectively; pD is be the prevalence of true disease.

prob[ M>c | D+ ] and prob[ M>c | D- ] are the sensitivity of marker vs. true disease at cut-off c and 1-specificity of marker vs. true disease at cutoff c.

For example, for a perfect marker panel where the sensitivity and specificity with respect to true disease are 100% (AUROC = 1), a disease prevalence of 40% and the sensitivity and specificity of the biopsy vs. true disease both equal to 90%, the observed sensitivity and specificity of the marker panel vs. biopsy will be equal to 0.86 and 0.93, respectively.

 

 

retour gH    Retour à la page principale de   (gH)