predict survival rate beyond follow up

[xinjian]

Dear All,

I have a longitudinal data set below:

ID	time	event	T
1	6	1	0
2	12	1	0
3	16	0	1
4	18	0	1
5	8	1	1

Total 60 subjects, maximal follow time is 24 months. Is it possible to use my dataset to predict survival rate for 5 or 10 years?

My program is below:

proc lifetest data = want plot= survival (nocensor failure test);

time time*event(0);

strata T;

run;

Thanks so much for your help,

Xinjian

[snoopy369]

It seems like it would be difficult to predict 5 or 10 year survival rates with any degree of precision with only 2 years of data, doesn't it? Some diseases have fairly stable 1-2 year survival curves with major drop-offs three or four years out, while some have fairly linear progressions. How does your model know if there's likely to be a spike in year 3 or 4?

[xinjian]

Hi,

I really appreciate your quick response. Yes, it is true to predict future 5 years using 2 year data set. I am more worried survival rate maybe flat. I just want know if there is a way to get survival rate beyond follow up window. I may use another dataset to validate the prediction.

Thanks,

Xinjian

[Reeza]

Not with any sort of accuracy. Survival curves are long tailed and where they flatten out/bend can be quite variable. 

[xinjian]

Dear Reeza,

Can you teach me how to write a program to predict survival rate beyond follow up 1, 2, or 3 years?

Thanks,

Xinjian

[Reeza]

If *I* had to do it, I guess I'd use PROC LIFEREG, where you'd build a model with parameters and once the model was built, you could likely extend it to a time beyond your data, BUT make sure to include the Prediction Intervals (not Confidence Intervals) to show how large your error will. 

I'm not even 100% sure that would work, but it's what I'd try.

[xinjian]

Many thanks for your suggestion. I'll try it.

[xinjian]

HI Reeza,

I use proc lifereg to build a model, the syntax is as follow:

proc lifereg data=want outest=expoutest;

model time*event(0)=T/dist=exponential;

title 'lifereg predict model';

output out=expsurv xbeta=exp_xb; run;

The estimates output is as follow:

Analysis of Maximum Likelihood Parameter Estimates
Parameter	DF	Estimate	Standard Error	95% Confidence Limits		Chi-Square	Pr > ChiSq
Intercept	1	4.1418	0.4472	3.2653	5.0184	85.77	<.0001
T	1	1.2695	0.6325	0.0300	2.5091	4.03	0.0447
Scale	0	1.0000	0.0000	1.0000	1.0000
Weibull Shape	0	1.0000	0.0000	1.0000	1.0000

My question is how can I write a syntax to set a time point at 3 year to predict survival rate?

Thank you!