Hello,

In this case the non-informative censorship is violated, and you have to work with an informative censorship models.

Best Wishes

Thank you. But my question remains as is there any standard cut off limit that till this point it is ok or like that (for both non informative or informative censoring).

For this point, it's up to you to check if a high % of loss to follow-up (LF) can change the results, il litterature , we found some researchers suggest a 5% of LF, others 10 % ...,

But in my opinion you can use a simulation of all this scenrio ( e.g: 1 % , 5 %, 10 % , 20 %) to check if these % affect your estimations.

Best luck dear.