How long does a knee replacement last? A systematic review and meta-analysis of case series and national registry reports with more than 15 years of follow-up

Key points

1. 01The primary outcome was all-cause survival of the implant construct, meaning the implant had not needed revision for any reason.
2. 02Registry data, which the authors trust more, gave 25-year survival of 82.3% for total knee replacement and 69.8% for unicondylar knee replacement.
3. 03Case series consistently reported higher (more optimistic) survival than registries, which the authors attribute to selection and publication bias.
4. 04No case series reported 25-year survival for total knee replacement, and the only case series 25-year unicondylar estimate (72.0%) rested on a single series.
5. 05The longest-term registry data (20 and 25 years) came only from Australia and Finland, limiting geographic generalizability.

How it was conducted

Design

Systematic review and meta-analysis of case series, cohort studies, and national joint replacement registry reports, following PRISMA (PROSPERO CRD42018105188)

Search

MEDLINE and Embase from inception to July 21, 2018, for primary TKR, UKR, and patellofemoral replacement series with mean or median follow-up over 15 years in osteoarthritis or unselected patients

Intervention

Total knee replacement (TKR), unicondylar knee replacement (UKR), and patellofemoral replacement; complex primary or revision procedures excluded

Outcomes

All-cause construct survivorship (no revision of any part) at 15, 20, and 25 years

Analysis

Fixed-effects meta-analysis weighting each series by its standard error; case series and registry data analyzed separately; study quality assessed with a non-summative four-point system

What they found

• 33 case series in 30 articles met inclusion: 6490 TKRs across 26 series and 742 UKRs across 7 series; no patellofemoral series qualified and no case series reported 25-year TKR survival.
• Registries contributed 299,291 TKRs (47 series) and 7714 UKRs (5 series), all from Australia and Finland for the longest timepoints.
• Pooled registry TKR survival: 93.0% (95% CI 92.8 to 93.1) at 15 years, 90.1% (89.7 to 90.4) at 20 years, and 82.3% (81.3 to 83.2) at 25 years.
• Pooled registry UKR survival: 76.5% (95% CI 75.2 to 77.7) at 15 years, 71.6% (69.6 to 73.6) at 20 years, and 69.8% (67.6 to 72.1) at 25 years.
• Case series were more optimistic, with pooled TKR survival 96.3% at 15 years and 94.8% at 20 years, and a single case series 25-year UKR estimate of 72.0% (95% CI 58.0 to 95.0).

Limitations

• Pooled estimates were not adjusted or stratified for patient factors such as age, sex, or indication, because individual patient data were unavailable.
• The longest-term registry data came only from Australia and Finland, with all UKR and 20 to 25 year TKR data from Finland alone, limiting geographic generalizability.
• Survival was assumed equivalent to risk and censoring (patients dying with the implant in place) was not accounted for, an assumption clearly violated in older cohorts.
• At 25 years the number of UKRs still at risk was under 25 in two registry series, so Kaplan-Meier accuracy at that timepoint is questionable.

Why it matters

For patients

If you are having a knee replacement, roughly 8 in 10 total replacements and 7 in 10 partial replacements are expected to still be working 25 years later.

For clinicians

Use the registry-based figures (about 82% TKR and 70% UKR survival at 25 years) rather than case series for patient counseling, resource planning, and consent, as case series overstate longevity.

For readers

This is the first generalizable estimate of knee replacement longevity, and it shows registry data give a more conservative and trustworthy picture than published case series.