Minimum follow up period survival analysis

Question

Imagin there's a surgery for a disease A that can cause a disease B. The minimum period to diagnose the disease B after the surgery of the disease A are 3 monthts (before that period of time, we can't say disease B is present, taking into account the literature published).

If I have a cohort of patients that have been operated for disease A and I want to study the incidence of disease B performing a survival analysis but some of them have been followed up less than 3 months (disease B didn't have the oportunity to show up), including them in the study doesn't make sense, right? I don't know if I should exclude them or since the survival analysis includes censored patients I should consider them as censored.

Answer 1

It won't matter, at least for simple Kaplan-Meier analyses or for Cox proportional hazards regression. Calculations for Kaplan-Meier and Cox analyses are based on the individuals at risk at times of events. There are two scenarios, depending on your choice of time = 0 reference point for the survival analysis.

If time = 0 is taken as the time of surgery for Disease A and it's impossible to detect Disease B until 3 months later, then those patients would have right-censored observations at times prior to 3 months. Thus they will never be in a risk set at an event time. They won't contribute anything to the calculations.

If time = 0 is taken as 3 months after the time of surgery for Disease A, then the individuals followed for less than 3 months after surgery aren't eligible for inclusion in the study at all.

On the off chance that Disease B might occasionally be detectable prior to that 3-month cutoff, I'd recommend using the first definition of time = 0 and including everyone. Whatever choice you make, however, no individual followed up for a shorter duration than the first observed event time contributes to these types of calculations.