New package kmsenspec on SSC

[Roger Newson]

Thanks as always to Kit Baum, a new package kmsenspec is now available for download from SSC. In Stata, use the ssc command to do this.

The kmsenspec package is described as below on my website. It comes with a manual kmsenspec.pdf, distributed as an ancillary file, giving the methods and formulas in detail.

Best wishes

Roger

-----------------------------------------------------------------------------------------------------------------------------------------------------------
package kmsenspec from https://www.rogernewsonresources.org.uk/stata16
-----------------------------------------------------------------------------------------------------------------------------------------------------------

TITLE
kmsenspec: Estimate sensitivity, specificity and predictive values from Kaplan-Meier curves

DESCRIPTION/AUTHOR(S)
kmsenspec is intended for use in a survival time dataset set up by stset.
It inputs a binary variable indicating a positive test result, and
estimates positive and negative predictive values from Kaplan-Meier
survival probabilities at a time specified by the user in positive and
negative observations, and then estimates the sensitivity and specificity
of the test using the Bayes theorem, with delta-Greenwood variances
estimated from the Greenwood standard errors of the positive and negative
predictive values. The estimation results are estimates of the
sensitivity, specificity, negative predictive power, and positive
predictive power, with a covariance matrix for the untransformed
estimates. Alternatively, kmsenspec can be used with the SSC packages
parmest and esetran to compute delta-Greenwood confidence intervals using
a variety of Normalizing transforms. The kmsenspec package uses the SSC
package kmest, which must be installed in order for kmsenspec to work.

Author: Roger Newson
Distribution-Date: 07 August 2025
Stata-Version: 16

INSTALLATION FILES (click here to install)
kmsenspec.ado
kmsenspec_p.ado
kmsenspec.sthlp

ANCILLARY FILES (click here to get)
kmsenspec.pdf
-----------------------------------------------------------------------------------------------------------------------------------------------------------
(click here to return to the previous screen)

[Dave Airey]

Roger Newson, do you use your package for survival model predictive accuracy and ROC curve analysis? (Survival model predictive accuracy and ROC curves - PubMed, Time-dependent ROC curves for censored survival data and a diagnostic marker - PubMed)

[Roger Newson]

Yes, the kmsenspec is used to compute sensitivity and specificity, but only if the PPV and NPV are estimated from a Kaplan-Meier curve. This might happen if (likre me) the user is testing an epigenomic diagnostic test for predicting cancer diagnoses in the following yrar, and not all trial subjects necessarily have the full year of follow up, as they were subjected to the diagnostic test less than a year ago, and we do not want to throw away the information from those patients. And this will be a valid thing to do, if the event of censorship is unrelated to the diagnosis status they would have had, had they not been censored.

The kmsenspec package comes with a manual kmsenspec.pdf, distributed on SSC as an ancillary file. The methods and formulas are explained in detail there.

I hope this helps. Let me know if you have any more queries.

Best wishes

Roger