Robotic knee replacement gave no patient benefit over standard surgery

Robotic arms are increasingly marketed as the future of joint replacement, promising more precise implant positioning and, by implication, better-feeling knees. A rigorous British trial designed to test that promise has now reported, and its verdict is blunt: for patients having a total knee replacement, a robotic-assisted operation delivered no clinically meaningful benefit over conventional instruments a year later — and it cost more [s1].

What the trial did

RACER-Knee was a pragmatic, superiority randomised controlled trial at ten hospitals in Great Britain, involving 33 surgeons [s1]. Patients with advanced knee osteoarthritis were randomly assigned 1:1 to have their knee replaced either with conventional instruments or with the Mako robotic-arm-assisted system [s1]. Crucially, this was a masked trial: patients and outcome assessors were kept unaware of which operation had been performed, using sham incisions, additional draping, and masked operation notes [s1]. That blinding matters enormously here, because expectations about high-tech surgery can inflate how patients report their own recovery — the exact bias a sham-controlled design is built to strip out.

The primary outcome was the Forgotten Joint Score at 12 months, a 0-to-100 patient-reported measure of how "aware" someone is of their artificial joint in daily life, where a higher score means the knee intrudes less on everyday activity [s1]. The trial was powered to detect a prespecified target difference of 12 points [s1].

What it found

Between December 21, 2021, and December 13, 2023, 807 patients were screened and 339 were randomly assigned — 168 to robotic surgery and 171 to conventional surgery [s1]. Their median age was 68.8 years (interquartile range 61.3 to 75.2); 167 (49%) were female and 172 (51%) were male [s1].

At 12 months, the mean Forgotten Joint Score was 49.2 (standard deviation 28.4; n=154) in the robotic group and 50.2 (29.9; n=158) in the conventional group [s1]. The adjusted mean difference was −1.5 points (95% confidence interval −7.5 to 4.5; P=0.62), a gap that not only fell far short of the 12-point target but actually pointed, trivially, in favour of conventional surgery [s1]. Safety was similar: 16 participants in each group had one serious adverse event [s1].

How to read it

This is close to a textbook null result, and its design makes it hard to explain away. Because patients were blinded with sham incisions, the score they reported reflects how the knee actually felt, not how impressed they were by the technology [s1]. Because the trial was pragmatic and spread across ten hospitals and 33 surgeons, the finding reflects robotic surgery "as delivered in routine practice," not a single enthusiast's results [s1]. And because the confidence interval around the difference (−7.5 to 4.5) sits comfortably inside the 12-point target, the trial can reasonably rule out a benefit of the size that would matter to patients [s1].

The economics sharpen the conclusion. The trial's own interpretation is that robotic-assisted surgery was more costly than conventional surgery and did not provide a clinically meaningful patient benefit at 12 months, with similar harms [s1]. A more expensive operation that does not make patients feel better is a hard proposition to justify on the evidence presented here.

Two honest caveats. First, this is a 12-month result; long-term follow-up is ongoing, and it remains possible that more precise implant placement could pay off in durability or revision rates over many years — though that is a hypothesis, not a finding [s1]. Second, the trial studied one robotic system for one operation in patients who did not need complex implants, so it does not speak to every device or every difficult knee [s1].

Funding, again, cuts in favour of the result's credibility: RACER-Knee was funded by the UK National Institute for Health and Care Research's Health Technology Assessment Programme — a public body with no stake in selling robots — rather than by a device manufacturer [s1]. The trial was registered before it began [s2].

For a patient offered a "robotic" knee replacement, the practical translation is that the label alone is not a reason to expect a better-feeling knee at a year. This article describes research and is not medical advice; the right operation and surgeon for a given knee is a decision to make with an orthopaedic team.

What to watch

The unresolved question is the long game: whether robotic precision eventually shows up as fewer revisions or longer-lasting implants over five to ten years, which the continuing follow-up of RACER-Knee and similar trials is designed to answer [s1]. Until then, the burden of proof sits with the technology to show a benefit that patients can actually feel or that outlasts its extra cost.

Sources

• Robotic-arm-assisted versus conventional total knee replacement (RACER-Knee) — The Lancet, 20 August 2026
• RACER-Knee trial registration (ISRCTN27624068) — ISRCTN registry, registered 8 July 2020