Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

R: R: st: R: reporting time dependent covariates in cox regression


From   "Carlo Lazzaro" <carlo.lazzaro@tin.it>
To   <statalist@hsphsun2.harvard.edu>
Subject   R: R: st: R: reporting time dependent covariates in cox regression
Date   Fri, 4 Sep 2009 15:04:21 +0200

Dear Moleps,
Thanks a lot for the follow-up message.
Your solution (separate rows for fixed and time-varying results)in the same table) seems wise and easy to get, given a limited theoretical introduction, even for referees or readers with just a smattering of Cox regression.

Kind Regards,
Carlo
-----Messaggio originale-----
Da: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] Per conto di moleps islon
Inviato: venerdì 4 settembre 2009 12.35
A: statalist@hsphsun2.harvard.edu
Oggetto: Re: R: st: R: reporting time dependent covariates in cox regression

thanks. From what, Maarten Buis, has taught me that the interpretation
of the time-varying variable is not too hard. You multiply the "fixed"
variable with the time varying variable raised to the power of time
you're interested in, ie at time 5 (days,minutes,years-depending on
your axis) you'll have the hazard ratio of the
-rh-variable*-t-variable^5. So in other words a -t- variable <1
reduces the hazard ratio over time and a -t-var >1 increases the
hazard ratio over time. Since the end-point in my study is death I'm
not surprised to find that several of my covariates have -t-var <1.
Intuitively most factors will diminish in importance as death is
imposing.

What is harder to interpret is an interaction between two time-varying
covariates.

I think what I'll do is to create a table and report my time-varying
covariates in a separate row beneath the "fixed" effect in case of
significance.


Moleps



On Fri, Sep 4, 2009 at 11:29 AM, Carlo Lazzaro<carlo.lazzaro@tin.it> wrote:
> Dear Moleps,
> in the first part of my previous reply I misunderstood that, from the very start of your message, you focused on tvc and not on proportional hazard condition only.
>
> <Cleves is a good book, [...] they dont either give any clues as to how results with -tvc-should be presented.>
>
> I agree with you that more examples can help readers. Hopefully this might be improved in the next edition of this helpful textbook.
> However, at pag 169 of the revised edition (par 10.5.1), authors report a brief description of the output of
>
> . stcox protect, tvc(init_drug_level) texp(exp(-0.35*_t))
>
> and highlight that
>
> "The hazard ratio 0.8848 is now interpreted as those with higher drug levels in their bloodstreams have a lower risk of having a hip fracture"
>
> as far as the time-varying variable is concerned.
>
> As a temptative hint, I was wondering whether the same approach  - that is, splitting the Cox regression table to allow for constant-with-time variables (rh) and time-varying-variables (t) Hazard Ratios to be separately reported; describing in the Results section of your research report what causes your Hazard Ratios to differ - may match with your research aims.
>
> Sorry I cannot be more helpful.
>
> Kind Regards and All the Best for your research project,
> Carlo
> -----Messaggio originale-----
> Da: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] Per conto di moleps islon
> Inviato: venerdì 4 settembre 2009 9.43
> A: statalist@hsphsun2.harvard.edu
> Oggetto: Re: st: R: reporting time dependent covariates in cox regression
>
> Appreciate your efforts, Carlo. However Lindholt et al had only
> dichotomised before and after x yrs, not the -tvc-option in stata. I
> agree that Cleves is a good book, but a little short on the practical
> examples-and they dont either give any clues as to how results with
> -tvc-should be presented.
>
> Regards,
> M
>
>
> On Fri, Sep 4, 2009 at 7:59 AM, Carlo Lazzaro<carlo.lazzaro@tin.it> wrote:
>> Dear Moleps,
>> As far as a part of your query is concerned
>>
>> <I still havent even seen anyone report whether PH assum_p_tions have been
>> met, so I cant really find any articles to copy>.
>>
>> I would refer you to the following article:
>>
>> Jes S Lindholt, Svend Juul, Helge Fasting and Eskild W Henneberg
>> Screening for abdominal aortic aneurysms: single centre randomised
>> controlled trial. BMJ, doi:10.1136/bmj.38369.620162.82. (Particularly,
>> Statistical analyses paragraph).
>>
>> For further details on Survival Analysis topics, I will recommend you to
>> take a look at:
>> Cleves MA, Gould WG, Gutierrez R. An Introduction To Survival Analysis Using
>> Stata. Revised edition. College Station: StataPress, 2004; [ST] Stata
>> manual. Survival analysis and epidemiological table. Release 9
>>
>> Two other relevant contributors of the Statalist - Maarten Buis
>> (http://home.fsw.vu.nl/m.buis/)  and Stephen Jenkins
>> (http://www.iser.essex.ac.uk/teaching/degree/stephenj/ec968/index.php.)
>> published really interesting papers as well as teaching-notes on Survival
>> Analysis.
>>
>> HTH and Kind Regards,
>>
>> Carlo
>>
>> -----Messaggio originale-----
>> Da: owner-statalist@hsphsun2.harvard.edu
>> [mailto:owner-statalist@hsphsun2.harvard.edu] Per conto di moleps islon
>> Inviato: giovedě 3 settembre 2009 22.32
>> A: statalist@hsphsun2.harvard.edu
>> Oggetto: st: reporting time dependent covariates in cox regression
>>
>> Dear listers,
>> What is a good way to report time-dependent covariates in a
>> multi-variate cox table? In my field of research I still havent even
>> seen anyone report whether PH assumtions have been met, so I cant
>> really find any articles to copy.
>>
>> I´m thinking in the lines of reporting an extra line beneath the main
>> effect in parenthesis for the significant time-dependent variables in
>> a table. Anyone have any experience in this?
>>
>> Also-how would you interpret a significant interaction between age and
>> treatment that is also time-dependent?
>>
>>
>> Regards,
>> M
>>
>> *
>> *   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/
>
>
>

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index