C-statistic in censoring data

[DF Moreno]

Hi everyone,

I have analyzed a cancer dataset that contains risk factors and an outcome. However, I am dealing with many observations lost to follow-up and missing values.
I want to obtain the c-statistic (using the "estat concordance" after doing a multivariate stcox analysis), but I was told that the c-statistic can be biased in the presence of censoring data and recommended me to use the Uno c-statistic.

Stata mentions that the Gonen and Heller estimate is better in case of censoring data. Is there a package available in Stata to calculate the uno c-statistic? Or both methods are correct in this scenario?


Many thanks

[Carlo Lazzaro]

DF:
you're probably looking for the -gheller- statistics, which is available from -help stcox postestimation- (see related helpfile and Stata .pdf entry for further details).

[DF Moreno]

Dear Carlo,

Many thanks. By the way, which is better to use: the Uno or the Gonen/Heller c-statistic?

[Carlo Lazzaro]

DF:
sorry, buy I cannot help you out with that.