0
$\begingroup$

I am working with a data that is I believe left truncated. I am looking at how Veterans with MPNs have a different timing of arterial thrombosis (AT) compared with Civilians with MPNs.

In the data some veterans and civilians have AT after being diagnosed with MPN, which is no problem. However, some of the veterans and civilians experienced AT before being diagnosed with MPN. We are only considering first AT. This could lead to immortal time bias.

From the nature of the question, it seems to me that patients become at risk when they are diagnosed with MPN. So patients who experience AT prior do not enter the risk group before they are diagnosed, however they can never experience first AT again.

Are the patients who experience AT first left truncated? I am planning on doing a log rank test for veterans and civilians, but should I remove these left truncated patients first?

Any guidance is appreciated!

$\endgroup$
2
  • $\begingroup$ What's an MPN? Also, you wrote "From the nature of the question, it seems to me that patients become at risk when they are diagnosed with MPN." Why does it seem that way to you? The data you present contradicts that. $\endgroup$
    – Peter Flom
    Commented Jul 30, 2023 at 11:33
  • $\begingroup$ MPN is myeloproliferative neoplasms. It seems that way to me because the research question focuses on veterans/civilians with MPN and timing of AT among these people. Is my interpretation wrong? $\endgroup$
    – Dnz857
    Commented Jul 30, 2023 at 11:41

1 Answer 1

1
$\begingroup$

In a comment you wrote:

the research question focuses on veterans/civilians with MPN and timing of AT among these people

In that case, people who got AT first are not part of your study sample and should not be included. And you should specify this in the write-up, under "exclusion criteria".

$\endgroup$
2
  • $\begingroup$ Thank you, I understand! I want to double check one more thing. If we are looking how Veterans with MPNs have a different timing of AT compared with Civilians with MPNs, more specifically if we are looking at when AT forms relative to MPN, before or after, would that require me to use competing risk analysis or how would I go about that? $\endgroup$
    – Dnz857
    Commented Jul 30, 2023 at 12:49
  • $\begingroup$ Well, you could make a variable of the difference in time between one and the other (this could be negative or positive) among people who had both. Then do separate analysis of people who have only one (that would be two survival analyses). $\endgroup$
    – Peter Flom
    Commented Jul 30, 2023 at 13:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.