Purpose: The aim of this study was to develop a more accurate method to deal with patients lost to follow-up based on the competing risks approach.
Methods: A cohort of 112 patients who received 143 primary cemented total knee arthroplasties forms the basis for this study. Follow-up was up to 25 years. The new method for dealing with lost to follow-up accounts for competing events (i.e. death and failure of a prosthesis) using the cumulative incidence estimator and estimates time to event for patients lost to follow-up using national demographic registries. The results of this new method were compared with the worst case scenario estimated by Kaplan-Meier.
Results: Six different situations were identified covering all possible situations in long-term follow-up for total knee arthroplasty. The new method--considering all patients lost to follow-up as revised--showed a twofold reduction in revision rate compared to the traditional worst case scenario using Kaplan-Meier.
Conclusions: Lost to follow-up should be prevented whenever possible, but this may be unavoidable for long-term follow-up studies. In situations where lost to follow-up does occur, the new proposed method offers an efficient and valid approach to deal with this problem.