Re: st: Left truncation in survival: how to use stset origin and enter

[Steven Samuels <ssamuels@albany.edu>]

Sacrificial Line-ignore.

On May 11, 2007, at 10:59 AM, Svend Juul wrote:
..
> I don't think this should be considered a case of delayed
> entry: The prognostic value of the lab examination applies
> to the time after the examination, and the date of diagnosis
> is hardly relevant in that context. So you could:
>
>    generate risktime = enddate - examdate
>    stset risktime , failure(...)
I disagree with Svend.  In the comparative trials that I have  
studied, the date of an examination would rarely, by itself, be a  
natural zero for the start-time. The exceptions that I can think of  
are: 1) The exam is triggered by an objective external event, such as  
the first occurrence of certain symptoms; in essence there is a new  
disease stage. 2) The exam determines eligibility for a new treatment  
or clinical trial.

Otherwise, the exam is unrelated to the course of the disease and so  
the exam date is not a natural start time. Consider--what if there  
had been more than one examination? Which should be chosen?

The exam provides information on time-dependent covariates,  
specifically, "internal time-dependent covariates" (Kalbfleisch and  
Prentice, 1978, The Statistical Analysis of Failure Time Data). These  
are often correlates of disease progression, for example CD4 counts  
in patients with HIV. They can have great prognostic value, which is  
what Diego seeks to ascertain.

This seems to leave date of diagnosis as the origin. However, without  
more information, I cannot tell what -stset- commands are needed. I  
need to know: Are the study patients part of a larger group, some of  
whom died prior to the exam or who did not receive the exam for other  
reasons?  Does Diego know the dates of diagnosis for the patients  
without exam?

I would also want to know more about the disease and how it was  
diagnosed.  Was it discovered:  by screening asymptomatic patients?  
2) during investigation of an unrelated illness? 3) because of  
symptoms directly or indirectly caused by the disease?

Disease diagnosed in asymptomatic patients is obviously at a very  
different stage than that diagnosed because of symptoms.

Steven



> Diego wrote:
>
> I have a question regarding the use of origin and enter
> option when stsetting data for a survival analysis with
> left truncation (delayed entry).
>
> I have the diagnosis date and the exam date for all my
> patients. The exam date can be also few years after the
> diagnosis date.
>
> Obviously the diagnosis date is the date patients start
> to be at risk (end-point: death) but the aim of the
> study is to look at the prognostic value of the lab
> examination we do on our study population, and the follow-
> up in that case is only 3 years (where the follow-up
> from the diagnosis time is 7-8 years).
>
> How should I stset my data to take into account the
> diagnosis date but to look at the prognostic value of
> the exam date ?
>
> ------------------------------------------------------------
>
> I don't think this should be considered a case of delayed
> entry: The prognostic value of the lab examination applies
> to the time after the examination, and the date of diagnosis
> is hardly relevant in that context. So you could:
>
>    generate risktime = enddate - examdate
>    stset risktime , failure(...)
>
> Hope this helps
> Svend
>
> ________________________________________________________
>
> Svend Juul
> Institut for Folkesundhed, Afdeling for Epidemiologi
> (Institute of Public Health, Department of Epidemiology)
> Vennelyst Boulevard 6
> DK-8000 Aarhus C,  Denmark
> Phone, work:  +45 8942 6090
> Phone, home:  +45 8693 7796
> Fax:          +45 8613 1580
> E-mail:       sj@soci.au.dk
> _________________________________________________________
>
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/