0
$\begingroup$

I'm a medical student performing my final degree thesis, so sorry in advance for my little knowledge in stats.

I'm performing a retrospective study on the surgical field where I want to evaluate the development of a anal fistula after the debridement of an anal abscess. At first, I thought about doing a survival analysis (and I even started to do so). However, I don't know if it makes sense, since when the wound is cured (a fistula is not formed), the patients are generally discharged from surgical controls. This cured patients have a little follow-up period compared to the ones where symptoms persist and are eventually diagnosed with a fistula. This generates a lot of early right-censoring in my Kaplan-Meier plots, and I believe that the results on my Cox regression model are biased because of that.

It would make more sense If I used Chi squared, Student T or Mann-Whitney-U tests for the univariant analysis and binary logistic regression for the multivariant analysis?

Thanks in advance!

Improve this question
$\endgroup$
3
  • $\begingroup$ Why do you think it would be biased? $\endgroup$
    – Lukas Lohse
    Commented Feb 12 at 22:32
  • $\begingroup$ @LukasLohse because most of the patients are early censored (the wound is cured and they're discharged at 3-6 months post-surgery, so as I understand, they contribute poorly to the analysis). The ones that are censored later are the patients that have recurrences of the abscess and even though the fistula might be present, it is not identified. So the ones that contribute most to the analysis are the ones that might have the chronic disease. $\endgroup$
    – user406570
    Commented Feb 13 at 22:20
  • $\begingroup$ Is your concern about the Kaplan Meier, i.e. the estimated probability of a fistula being to high or about the coefficients in the Cox-model being off? $\endgroup$
    – Lukas Lohse
    Commented Feb 14 at 15:53

0

Reset to default

Your Answer

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