Hi everyone, I hope you guys can help with this - I would be very very grateful!

I am quite comfortable with standard Cox-regression method, but I've always struggled with time-dependent covariate analysis.

Basically, I want to identify variables that are associated with the development of cancer outcome.

This is a right-censored data with patients leaving the study when they develop cancer, or lost in follow/up etc. Starting point is patient entering the endoscopic follow-up.

There are 7 variables, all of which are binary (please see attached for the current data format). Once patient develop one of these variables, they cannot go back to the "negative for risk factor" group.

But my problem is, each patient can develop any of these risk factors any time between the start till the end of follow-up (i.e. all of these are time-dependent covariate). For example, patient may be positive for risk factor 1 on entering the study, but may develop risk factor 2 in 5 years, and risk factor 3 in 2 years...etc.

I've come to realize that the standard cox-regression would probably not valid, as patient's risk factors profile changes over time.

I think this probably require some sort of time dependent covariate cox regression, but not quite sure how the data should be entered (which I am happy to put in more work as necessary) and once appropriate format of data is collected how this should be analyzed in SPSS. I note that there is a time-dependent analysis option with T_ variable, but since there are more than 1 time-dependent factors I am not so sure I this should be done.

Binary logistic-regression is tempting, but I realize that this is probably not valid since this does not really take into account "follow-up" time hence not appropriate for censored data.

Any piece of advice would greatly be appreciated!!

Many thanks,

BW

All of these variables are all binary - i.e. are 1 or 0.