Will the Knee Outlive the Patient? Kaplan-Meier Systematically Overstates the Lifetime Risk of Revision After Total Knee Arthroplasty.

BACKGROUND

Lifetime revision risk after total knee arthroplasty (TKA) is usually estimated from Kaplan-Meier survivorship, which treats death as noninformative censoring and overstates revision incidence when competing mortality is substantial. The bias increases with age. We quantified the discrepancy between Kaplan-Meier and competing-risk estimates and derived the age- and sex-specific probability of dying with the original TKA.

METHODS

We analyzed 2,920 primary TKAs from a 15-year multisurgeon single-institution database linked to the New Zealand Joint Registry. Revision and death were modelled as competing events using Aalen-Johansen cumulative incidence functions and Fine-Gray subdistribution hazards, with patient-level clustering for bilateral cases. Kaplan-Meier estimates censoring death served as the benchmark. Cohort mortality was compared with the New Zealand population using standardized mortality ratios (SMRs) from 2017 to 2019 life tables.

RESULTS

There were 155 revisions and 525 deaths (median follow-up 7.4 years). The 15-year revision estimate was 8.3% by Kaplan-Meier versus 7.4% by competing-risk estimation overall, and 5.6% versus 3.0% in patients aged 85 years or older (relative overestimation >80%). At 10 years, cumulative incidence of revision was 5.7% (95% CI 4.8-6.7) versus 19.9% (18.2-21.8) for death, diverging with age (85 years or older: 3.0% versus 67.6%). 97% of patients aged 75 years or older died with the original TKA. Each 10-year age increment lowered the revision hazard (sHR 0.76, 0.61-0.94) and raised mortality (HR 2.33, 2.02-2.68). Cohort mortality was lower than the general population (SMR 0.73, 0.66-0.79). Among Māori, this relative advantage (SMR 0.66) was absent against the total population (SMR 1.11).

CONCLUSIONS

Kaplan-Meier systematically overstates absolute lifetime revision incidence after TKA, most in older patients for whom competing mortality dominates. Age- and sex-specific competing-risk estimates provide more accurate preoperative figures and should be reported alongside Kaplan-Meier when presenting long-term revision risk.

LEVEL OF EVIDENCE: Prognostic Level III. See Instructions for Authors for a complete description of levels of evidence.