# Thread: Kaplan Meier Curves + Cox PH

1. ## Kaplan Meier Curves + Cox PH

Hi all:

I'm doing a survival analysis on a dataset of about 650 cancer patients. The outcome is time to death. Patients are censored when they are (a) lost to follow up or (b) the study ends, which was Dec 31, 2013. I'm investigating the kaplan meier curves and I notice that for each covariate, the curves intersect. From what I gather, this means that Cox Proportional Hazards may be in appropriate. I've began to investigate some parametric models (the exponential, weibull, log-logistic, log-normal) and so far, none of them seem to fit. Here are some additional details about the dataset and disease.

-650 patients
-21 deaths (events) in total
-Follow up time (years) ranges from a few months (1-2 cases) to 25 years+ (50+ patients)

What are my options given that the kaplan meier curves intersect in terms of proceeding further? We would like to fit a multivariable model of some sort to adjust for patient risk factors.

TIA

2. ## Re: Kaplan Meier Curves + Cox PH

Even though the Kaplan-Meier curves intersect, this doesn't necessarily mean the proportional hazards assumption doesn't hold. UCLA has a pretty clear description of the different methods to use to check the proportional hazards assumption, with example using most of the popular statistical packages: http://www.ats.ucla.edu/stat/example...rtionality.htm

3. ## Re: Kaplan Meier Curves + Cox PH

Originally Posted by Disvengeance
Even though the Kaplan-Meier curves intersect, this doesn't necessarily mean the proportional hazards assumption doesn't hold. UCLA has a pretty clear description of the different methods to use to check the proportional hazards assumption, with example using most of the popular statistical packages: http://www.ats.ucla.edu/stat/example...rtionality.htm
Thanks for the reply + link. We're not going to go ahead with the regression modeling as the event rate is too low. Going to keep it descriptive

4. ## Re: Kaplan Meier Curves + Cox PH

In the past, I thought I interpreted crossing lines as differences in rates if the pull away from each other. Or if they end up intertwining, then equality of rates or indedence.

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts