The association between changes in a biomarker and time-to-event

Nobuyuki — Tue, 19 Oct 2021 00:02:03 GMT

Hi.

I'm interested in a biomarker.

The biomarker have been shown to be associated with worse outcomes (time-to-event) (the higher the biomarker is, the worse prognosis).

However, some therapeutic interventions increase the biomarker levels.

So, I suspect that the biomarker is beneficial, and the increase in the biomarker levels indicates better outcomes.

Here, we have dataset including repeated measurements of the biomarker and other variables at 2-3 time points (at baseline, 1-year, and 2-year).

Also, the dataset have information about time-to-event.

Time-dependent cox regression model is not appropriate, because it only showed that "the higher, the worse".

It does not provide me "what the changes in the biomarker affect the outcomes".

I'd like to know how I should analyze and how to do it with SAS.

Any comments would be appreciated.

Best regards,

Nobuyuki

Re: The association between changes in a biomarker and time-to-event

japelin — Tue, 19 Oct 2021 01:43:05 GMT

Please don't multi-post.

Nobuyuki — Tue, 19 Oct 2021 02:17:12 GMT

You mean post at Japan SAS discussion?

I just want comments both from international colleagues and Japanese colleagues.

But if it is inappropriate, I will delete the Japanese one.