Dr Matthew BroadheadOrthopaedic Surgeon — Sydney & Coffs Harbour

Journal Club · September 2026

Does robotic-assisted knee replacement improve your outcome?

A 2023 review of 12 randomised trials found robotic-assisted knee replacement placed implants more accurately than conventional surgery, but that accuracy did not translate into a meaningfully better pain-and-function score.

Robotic-assisted knee replacement is heavily marketed, and patients often ask about it directly. If a robot can position the implant more precisely, will my knee replacement work better? This 2023 systematic review pooled every randomised trial that has tested that question directly, and its answer is a useful corrective to how the technology is often sold.

What the study looked at

The researchers searched four databases for every published randomised controlled trial comparing robotic-assisted total knee replacement (RATKA) with conventional, jig-based total knee replacement (COTKA), in adults having surgery for primary knee osteoarthritis. They found 12 trials, run in several countries and using different robotic systems (ROBODOC, the oldest and most-studied system, was the most common; NAVIO and Mako, closer to what is used today, appeared in others). Together the trials covered 2,200 patients.

The review pooled two very different kinds of outcome: how accurately the implant was positioned on post-operative imaging (deviation from the planned mechanical alignment, and the rate of alignment “outliers”), and how the patient actually did (validated pain-and-function scores such as the WOMAC index, knee range of motion, complications, and revision surgery).

What they found

  • Accuracy: robotic assistance was measurably better. Alignment “outliers”, meaning knees ending up outside the target range, fell from about 125 per 1,000 with conventional surgery to about 53 per 1,000 with robotic assistance (risk ratio 0.43, 95% CI 0.27–0.67). Average alignment sat 0.94° closer to neutral (95% CI –1.1° to –0.73°).
  • But the authors question how much even that accuracy matters. This is the part most summaries leave out. They note the outlier finding translates to a risk difference of only 22 per 1,000, and a number needed to treat of 45, meaning 45 knees would have to be done robotically to avoid one extra outlier, which they say “may be considered non-clinically important”. They go further and question whether the sub-1° difference in average alignment is important at all, given that the threshold usually taken to matter for this measure is 3°.
  • Patient-reported outcome: no meaningful difference. The WOMAC pain-and-function score differed by 0.35 points (95% CI –0.78 to 0.07), against a threshold of 15 points for a difference a patient would actually notice. The gap is roughly forty times smaller than the amount that would register.
  • Range of motion: no clear difference (0.73°, 95% CI –7.5° to 6.0°, wide and inconclusive, against a 30° threshold for importance).
  • Revision and major complications: no difference detected, though only one of the twelve trials followed patients long enough to compare implant survival at 15 years (98% survivorship in both groups).

What it means for you

The honest summary is that robotic assistance does what it is designed to do: it places the implant closer to the surgical plan, more consistently, than a human working with jigs and cutting guides alone. What it has not yet been shown to do, across the best available randomised evidence, is make the knee feel or function better for the patient. That is why this practice describes robotic assistance as a tool used where it adds value, rather than as something that by itself produces a better result. If you are weighing up a robotic versus a conventional approach, this is useful background for that conversation with your surgeon, not a reason to insist on, or avoid, either.

A caveat worth knowing

Several things could still change this picture. The trials are dominated by an older robotic system (ROBODOC) that is not what most surgeons use today, follow-up was mostly short (months rather than years), and every trial carried an unavoidable risk of bias because the operating surgeon could not be blinded to which technique they were using. Larger trials using current robotic platforms and following patients for two years or more are still reporting, and may yet show a clinical benefit that this generation of evidence was underpowered to detect. For now, “more accurate” and “better outcome” are not the same claim, and the second has not been established.

Assessing this studyA standardised appraisal for clinicians and registrars: study design, strength and validity.
Study design
Systematic review and meta-analysis of 12 randomised controlled trials comparing robotic-assisted (RATKA) with conventional, jig-based (COTKA) total knee replacement, a therapy question.
Level of evidence
OCEBM Level 1 (therapy), the highest evidence tier for this comparison; even so, most included trials carried an unavoidable high risk of performance bias, and fewer than half had been registered.
Are the results valid?
12 RCTs, 2,200 patients, assessed with the Cochrane RoB 2.0 tool. Every trial had a high risk of performance bias because the operating surgeon could not be blinded to the technique used; allocation concealment and outcome-assessor blinding were often unreported, and only 5 of the 12 trials had been registered, so reporting bias could not be assessed in the rest. GRADE certainty was moderate for the alignment outcomes and the WOMAC score, low for range of motion, and long-term data were sparse (only one trial reported implant survivorship, at 15 years).
What are the results?
Accuracy: robotic assistance nearly halved the risk of a mechanical-alignment outlier (risk ratio 0.43, 95% CI 0.27-0.67; number needed to treat 45) and shifted mean alignment 0.94 degrees closer to neutral (95% CI -1.1 to -0.73, moderate certainty). Function: the WOMAC score, the main patient-reported outcome, differed by only 0.35 points (95% CI -0.78 to 0.07, moderate certainty), against a minimal important difference of 15 points, so roughly forty times smaller than a patient would notice; range of motion differed by 0.73 degrees (95% CI -7.5 to 6.0, low certainty) against a threshold of 30 degrees. Notably the authors also question the clinical importance of the accuracy gain itself, observing that the outlier result equates to a risk difference of only 22 per 1,000 and a number needed to treat of 45, and that the sub-1-degree alignment difference falls below the 3-degree threshold usually taken to matter. No difference was found in revision rate or major complications, though only one trial followed patients long-term (15-year survivorship, 98% in both groups).
Do they apply to our patients?
Applicable to adults having primary total knee replacement for osteoarthritis, the population any TKA surgeon sees day to day. It does not resolve which robotic platform matters: ROBODOC, the most common system across the included trials, differs from the imageless systems in current use. Most trials also followed patients for months rather than years, so it cannot speak to implant survival with modern systems.
Bottom line
Robotic assistance measurably improves how precisely the implant is positioned, but across the best available randomised evidence it has not been shown to make a clinically meaningful difference to how the knee feels or functions, and the review's own authors question whether the size of the accuracy gain is clinically important either.

Framework: study design graded with the Oxford Centre for Evidence-Based Medicine (OCEBM) Levels of Evidence; validity and applicability appraised in the CASP tradition.

This is a plain-language summary of published research, provided for general education. It is not medical advice and does not describe Dr Broadhead's own results. Whether any finding applies to you depends on your individual circumstances, so please discuss your care with your treating team.

← All Journal Club entries

CallEmail