Aims. The aims were to assess whether joint-specific outcome after total knee arthroplasty (TKA) was influenced by implant design over a 12-year follow-up period, and whether patient-related factors were associated with loss to follow-up and mortality risk. Methods. Long-term follow-up of a randomized controlled trial was undertaken. A total of 212 patients were allocated a Triathlon or a Kinemax TKA. Patients were assessed preoperatively, and one, three, eight, and 12 years postoperatively using the Oxford Knee Score (OKS). Reasons for patient lost to follow-up, mortality, and revision were recorded. Results. A total of 94 patients completed 12-year functional follow-up (62 females, mean age 66 years (43 to 82) at index surgery). There was a clinically significantly greater improvement in the OKS at one year (mean difference (MD) 3.0 (95% CI 0.4 to 5.7); p = 0.027) and three years (MD 4.7 (95% CI 1.9 to 7.5); p = 0.001) for the Triathlon group, but no differences were observed at eight (p = 0.331) or 12 years’ (p = 0.181) follow-up. When assessing the OKS in the patients surviving to 12 years, the Triathlon group had a clinically significantly greater improvement in the OKS (marginal mean 3.8 (95% CI 0.2 to 7.4); p = 0.040). Loss to functional follow-up (53%, n = 109/204) was independently associated with older age (p = 0.001). Patient mortality was the major reason (56.4%, n = 62/110) for loss to follow-up. Older age (p < 0.001) and worse preoperative OKS (p = 0.043) were independently associated with increased mortality risk. An age at time of surgery of ≥ 72 years was 75% sensitive and 74% specific for predicting mortality with an area under the curve of 78.1% (95% CI 70.9 to 85.3; p < 0.001). Conclusion. The Triathlon TKA was associated with clinically meaningful greater improvement in knee-specific outcome when compared to the Kinemax TKA. Loss to follow-up at 12 years was a limitation, and studies planning longer-term functional assessment could limit their cohort to patients aged under 72 years. Cite this article: Bone Jt Open 2024;5(10):911–919