SV: st: Relative survival using matched controls

[Katja Maretty Nielsen <KATJA.MARETTY.NIELSEN@ki.au.dk>]

Dear Adam

Many thanks.

I've tried both the suggested methods and decided on using the first strategy.

It seemed slightly more intuitive to me, especially since I didn't have proportional hazards.


Katja Maretty Nielsen


________________________________________
Fra: owner-statalist@hsphsun2.harvard.edu [owner-statalist@hsphsun2.harvard.edu] på vegne af Adam Olszewski [adam.olszewski@gmail.com]
Sendt: 19. august 2013 14:55
Til: statalist@hsphsun2.harvard.edu
Emne: Re: st: Relative survival using matched controls

Hi Katia,
You can make a long table and plot the estimates using the -strs- command or (easier and more elegantly, if you can live without the step curve) plot the relative survival using predictions (usually perfect as long as your N is reasonable) from a flexible parametric model fitted using -stpm2-
stpm2 is available from SSC and strs from pauldickman.com
Hope that helps,
Adam

Sent from my iPhone

On Aug 19, 2013, at 6:45 AM, Katja Maretty Nielsen <KATJA.MARETTY.NIELSEN@ki.au.dk> wrote:

> Dear all
>
> I'm trying to figure out how to calculate relative survival in cancer patients where I have the "expected survival" as a general population cohort of individually matched controls (5 controls for each cancer patient matched on age and sex at the date of cancer diagnosis).
>
> I have a data set containing the following variables:
>
> idno            identification number, unique to each individual
> age             age at diagnosis of cancer/index date
> gender
> type            type of individual. 1 Cancer patient   2 control
> group          number correlating each cancer patient with his/hers 5 individually matched controls
> foll              follow up time (years) since diagnosis/indexdate
> status         status at end of follow up.   1 dead    2   alive/censored
> com           comorbidity according to the charlson index categorized in 3 levels; no, mild, moderate, and severe comorbidity
> diayear       year of diagnosis/index year
>
> + some additional prognostic variables for the cancer patients such as stage at diagnosis, tumor size, location of primary tumor etc.
>
> What I would like to do is to produce Kaplan Meier curves for the relative survival, both as overall and for subgroups, e.g., adult patients with non-metastatic disease etc.
>
> I have looked into some of the command such as -strel- and -stexpect-, however these methods seems to be based on population mortality rate tables.
>
> I'm currently using stata version 11.2
>
> Many thanks
>
> Katja Nielsen
>
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/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/faqs/resources/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/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/