Medical vs. Epidemiologic vs Biostatistical question | Coding dates for untimely events

[Girish Venkataraman]

I am looking at a cohort of high-grade lymphoma transformed from a low grade lymphoma with predictor variables related to low-grade lymphoma and overall survival after development of transformed lymphoma. I have a few cases where low-grade lymphoma was detected concurrently as the high grade or even worse, a year after high-grade was detected.

How do I code the dates of low-grade and date of high grade in something like this where I am looking at time to transformation after low grade diagnosis and time to event (death) after transformation?

I don't want to drop the observation for sure. Generally accepted consensus on concurrent low-grade/transformation is up to within 3 months preceding date of transformation. But dates need to be sequential for analysis.

[Nate Tillern]

I don't know anything about the actual medical topic, but is there any reasonable explanation for why a low grade diagnosis occurs after a high grade diagnosis? I assume there is if you don't want to drop the observations, but if there is no sensible explanation for why the diagnosis occurred this way then in all likelihood there was human error in data entry.

You could simply have a negative value (or zero when concurrent) for "time to transformation" equal to the time between high grade and the later low grade diagnosis multiplied by negative one. While this should work mathematically, I feel like it would not work from a theoretical standpoint.

[Girish Venkataraman]

This phenomenon frequently happen in medical literature. Low-grade cancer is not symptomatic all the time but high grade cancer that arises from it is always symptomatic. A biopsy tissue of the high grade cancer may not necessarily harbor the low grade component that may become apparent (by radiologic imaging and or another biopsy) after the high grade cancer is treated. Biologically, the low grade cancer should have been there all along preceding the high grade. Just that I would forced to take a guess, but it will affect the analysis depending on what I choose. A date within 3 months preceding the high grade is what I was thinking but I don't know if that is correct. I will discuss your option with the clinical team and see if that seems reasonable.