1
$\begingroup$

I am relatively new to survival analysis and have seen survival outcomes often being reported using both progression-free survival/overall survival Kaplan-Meier curves in months, as well as hazard ratios in clinical papers. Is there a specific reason for this? Isn't it sufficient to report only one type of statistical result – ideally the finding in months which is more useful in clinical conversations?

$\endgroup$
1
  • $\begingroup$ From my experience, a lot of people confuse adjusted survival curves from a Cox model as Kaplan Meir curves. I've had to correct people on multiple occasions $\endgroup$ Commented Apr 28, 2021 at 19:19

1 Answer 1

0
$\begingroup$

Kaplan-Meier curves display something close to raw data: survival over time perhaps broken down into a few groups, but with no further adjustment for other variables. If you are just doing a Cox model to differentiate those same few groups, then the hazard ratios do provide quantitative estimates of the risks that some people (like me) want to see with the survival plots, plus a compact display of the standard errors of those estimates. With more than a couple of Kaplan-Meier curves, superimposing error estimates onto the curves can get confusing very quickly.

Cox modeling can be much more useful than Kaplan-Meier curves when there are multiple variables associated with outcome. Not only would the number of curves to display become unwieldy, but subgroups broken down by multiple variables soon become very small, diminishing the power to find true differences in outcome. Cox models allow for sharing of information among all cases, providing an efficient way to estimate associations of all the variables with outcome and to make survival predictions for any combination of variable values. The survival curves you get from the Cox model are predictions, not raw data, so keep that in mind (as Demetri Pananos said in a comment).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.