# Re: st: AGE at event occurrence as the time variable in mortality analysis rather than time between entry into follow up an event occurrence

 From Steven Samuels To statalist@hsphsun2.harvard.edu Subject Re: st: AGE at event occurrence as the time variable in mortality analysis rather than time between entry into follow up an event occurrence Date Sun, 7 Sep 2008 14:46:49 -0400

1. Svend and Maarten's advice for setting age as the time variable is still good.

2. You could add time from entry as a time-dependent covariate. However I suggest that you enter it in the form Z = age at entry. If Y is current age, then Y - Z is time from entry: the three are collinear. Age at entry is also equivalent to year of birth. Expressed in that way, it might capture some interesting effects of birth during wartime.

3. The risk sets of long-term survivors may be different genetically from the earlier risk sets--hardier, for example. This is the problem of heterogeneity. I suggest you employ one of the frailty options in -stcox-.

-Steve

On Sep 7, 2008, at 1:17 PM, uri goldbourt wrote:

Thanks.

I may have provided an incomplete explanation of the issue.
I have now 42 year mortality follow up of a cohort of men who had been
tenured civil servants and municipal employees in 1962.
In earlier analyses I always worked with time between entry (all entered in
1963 and early 1964, spread over 14 month) and date of event. Even 23 yr
mortality we analyzed in this manner.

With this new follow up, where 8471 of the original 10059 men have died, I
thought it more sensible to replace time till event (from entry) by age (or
time since birth' as you say) as the time variable. But the original sample
is not after "weeding out the high risk men" because they were aged 40-65 at
entry. I suppose this seems to justify handling the stset situation ignoring
left censoring. I ask, technically, whether I define the time variable
separately or include it anywhere in the stset statement?

Left censoring does, however, certainly plays a central role when I limit
analysis to survivors by 2000 (when everyone had either died or reached age
80), because only for them we have information on dementia (mostly Alzheimer
disease) assessed on 2000 and we want to know how. for them. parameters
assessed long before, at mid-life [clinical, biochemical, anthropometric,
social] were related to the age at death between 2000 and 2006.

UG
------------------------------------
-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Maarten buis
Sent: Sunday, September 07, 2008 9:49 AM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: AGE at event occurrence as the time variable in mortality
analysis rather than time between entry into follow up an event occurrence

--- uri goldbourt <goldbu1@post.tau.ac.il> wrote:

```In an epidemiological study where we follow up very old individuals
the importance of time till event (say, mortality) pales in
comparison to the actual age of the individual when the endpoint
event occurred. <snip> What is the accurate way to stset the data so
that age at event(assuming of course it is included for each event)
become the time variable?
```
Setting time is the easy part of your analysis: The age is nothing
other than the time elapsed since birth. In your case the problem is
left censoring: People could have experienced the event before entering
the study. As a consequence you have ended up with a selected sample
where the weaklings have already been weeded out. I have never done a
study where I had to deal with left censoring, so I don't remember if
Stata can do anything about that and my books are at the office.

-- Maarten

-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

Buitenveldertselaan 3 (Metropolitan), room Z434

+31 20 5986715

http://home.fsw.vu.nl/m.buis/
-----------------------------------------

*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/

*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
```*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```