Basic question about survival analysis

[Bruno P Nunes]

I'm starting to do survival analysis and I am a basic problem.

I have this data structure (one observation per row):

. list failure datebegin faildate timey

+-----------------------------------------+
| failure datebegin faildate timey |
|-----------------------------------------|
1. | 0 20jun1983 . . |
2. | 1 10jun1992 04feb2007 5352 |
3. | 1 30may1993 . . |
4. | 1 12oct1993 05may2007 4953 |
5. | 0 14dec1993 . . |
|-----------------------------------------|


Failure is the outcome (death); faildate is the date of death; and timey is the difference between beginning of treatment and death. I don't have faildate for who didn't die.

I want perform a survival analysis considering failure 0 and 1. But when I declare dataset to be survival-time data (stset), I just can analyse who died.

How I can consider who lived (faliure "0")?

Thank you.

[Marcos Almeida]

Hello Bruno,

I gather your study must "end" somehow. I mean, for those without the event, "faildate" happens when the study reaches the end and, therefore, the surviving individuals are right censored.

Hope that helps,

Best,

Marcos

[Bruno P Nunes]

Many thanks Marcos

I trying to do this putting "31dec2013" in faildate for those who lived. But, for me, the Kaplan graph seems strange. Please, see the attachment

Is this procedure correct? Is this right censored?

Thank you in advance.

Best, Bruno.

[Carlo Lazzaro]

Bruno:
as Marcos have already pointed out, you can't have a faildate for those who did not fail; in fact, they're right-censored (i.e., the investigator can't say whah will happen to them after the study time is over).
Your KM appears strange at the beginning, whereas it accounts for right-censoring after the 30th (year?) of observation.
As an aside, I would strongly recommend you the following valuable textbook: http://www.stata.com/bookstore/survi...ion/index.html.

[Marcos Almeida]

Hello, Bruno,

As Carlo underlined, some of your doubts may be solved after a short introduction on the matter.

That said, with regards to the graph, maybe you don't have much follow-up information > 16 years in your data set.

You might check if there are lots of censored individuals since then. You can use the "at risk" option to produce a graph with the population at risk.

Finally, your median survival time will be around 9 years.You can easily check this (and extra information) by typing:

. stdescribe

Best,

Marcos

[Bruno P Nunes]

Many thanks to Carlo and Marcos

The highlights were very helpful.

Best, Bruno.