Robotic-assisted total knee replacement has become one of the most important technological developments in modern joint replacement. Advocates point to greater surgical precision, reproducibility, intraoperative data, and the ability to individualize implant positioning. Critics have continued to ask a simpler question: Do patients actually do better?
The newly published RACER-Knee trial provides some of the strongest randomized evidence yet addressing that question. Published in The Lancet in August 2026, RACER-Knee was a pragmatic, multicenter, participant- and assessor-masked randomized controlled trial comparing Mako robotic-arm-assisted total knee replacement with conventional instrumentation.
For Episode 41 of The Joint Replacement Podcast, I sat down with two of the investigators behind the trial, Professor Andy Metcalfe and Professor Ed Davis, to discuss not only what RACER found, but what those findings actually mean for surgeons, patients, hospitals, and the future of robotic knee replacement.
Watch Episode 41: RACER Knee Trial with Andy Metcalfe & Ed Davis
What Was the RACER Knee Trial?
RACER stands for Robotic Arthroplasty Clinical and cost Effectiveness Randomised controlled trial. The study was designed at a time when robotic-assisted knee replacement was expanding rapidly, while high-quality evidence demonstrating clinical and economic superiority remained limited. The original protocol specifically set out to determine whether robotic-assisted total knee replacement provides meaningful benefit to patients and whether that benefit justifies its additional cost.
The final trial included 339 patients across 10 hospitals in Great Britain and 33 surgeons. Patients were randomized to robotic-assisted TKR using the Mako robotic-arm system or conventional total knee replacement. Importantly, this was a superiority trial, meaning the investigators were looking for evidence that robotic surgery produced better outcomes rather than simply demonstrating that it was equivalent to conventional surgery.
The study also went to considerable lengths to minimize bias. Patients and outcome assessors were masked to treatment allocation, with techniques including additional draping, sham incisions corresponding to robotic tracker placement, and masked operative documentation.
The Headline Finding: More Technology Did Not Mean Better 12-Month Outcomes
The primary endpoint was the Forgotten Joint Score (FJS) at 12 months, with a prespecified target difference of 12 points.
At one year, the mean FJS was:
49.2 in the robotic group
50.2 in the conventional group
The adjusted mean difference was −1.5 points, with a 95% confidence interval of −7.5 to 4.5 and a p value of 0.62. In other words, RACER did not demonstrate a clinically meaningful improvement in the primary patient-reported outcome with robotic-assisted TKR at 12 months. Serious adverse events were also identical in number, occurring in 16 participants in each group.
That is an important result, and both Professor Metcalfe and Professor Davis were very clear about it during our conversation.
As Davis summarized it, patients should not expect to feel better at 12 months simply because their knee replacement was performed robotically. But that does not necessarily answer every question about the value of robotics.
RACER Demonstrated Precision. The Question Is What We Do With It.
One of the most interesting themes of our conversation was the distinction between precision and outcome.
Robotic systems can help surgeons execute a surgical plan with considerable precision. But the robot does not independently determine the optimal position for an individual patient’s knee replacement. That remains dependent on surgical planning, alignment philosophy, soft-tissue balancing, implant design, and ultimately surgeon decision-making.
Davis summarized this concept memorably during the episode: there is little value in being extremely precise if the target itself is wrong.
Metcalfe similarly argued that RACER can be interpreted in more than one way. One interpretation is that additional precision simply does not matter enough to influence patient outcomes. Another is that we now possess a more precise surgical tool but have not yet determined how best to use that precision to improve outcomes.
That distinction may ultimately prove more important than the simple question of “robot versus manual.”
Did Alignment Philosophy Matter?
This was one of the most interesting areas of discussion.
RACER was designed years before many of today’s conversations around functional alignment, restricted kinematic alignment, and increasingly individualized approaches to TKA had matured. The robotic procedures therefore do not necessarily represent how every high-volume robotic surgeon performs TKA today.
Metcalfe explained that the study generally began from a mechanical alignment plan while permitting surgeons to make adjustments. In practice, he characterized the robotic strategy as resembling a restricted functional alignment approach.
This raises an important question.
If the primary advantage of robotics is the ability to measure anatomy and soft-tissue balance and then individualize implant position, will simply executing a relatively traditional alignment philosophy more precisely ever produce dramatically different patient-reported outcomes?
We do not yet know.
Davis said that if he were designing RACER today, alignment philosophy would likely need to be incorporated differently. At the same time, he acknowledged the fundamental difficulty: we still do not know exactly which individualized alignment strategy should serve as the target.
What About Surgeon Experience?
Another criticism following publication involved the robotic experience of participating surgeons.
RACER intentionally represented a pragmatic healthcare environment rather than limiting participation to a handful of extremely high-volume robotic surgeons. Some surgeons had relatively limited robotic experience, while others had performed hundreds of robotic procedures before entering the trial.
Metcalfe told us that the investigators have examined outcomes across different levels of robotic experience and have not seen a clear signal suggesting that greater robotic experience would have fundamentally changed the primary result. A dedicated learning-curve analysis is also forthcoming.
This highlights an important distinction between efficacy and effectiveness. A technology may perform differently in the hands of a small group of expert users than when deployed across an entire healthcare system. RACER was intentionally designed as a pragmatic trial to address the latter question.
One of the Most Fascinating Findings: What Patients Thought They Received
RACER’s masking produced another intriguing observation.
During the episode, we discussed data showing that patients who believed they had undergone robotic surgery had an FJS of approximately 53, whereas those who believed they had received conventional surgery scored approximately 31.
That should not automatically be interpreted as proof of a placebo effect.
As Davis explained, causality could run in either direction. Patients may have felt better because they believed they received advanced technology, or patients who were already doing well may have concluded that they must have received robotic surgery.
Either way, the finding underscores why rigorous masking is particularly important when studying highly marketed surgical technologies.
What About Cementless Robotic Knee Replacement?
There is another important limitation when applying RACER to current practice, particularly in the United States.
The knees in RACER were cemented.
Davis specifically confirmed during our discussion that the trial cannot answer whether robotic assistance offers different advantages in cementless TKA.
That distinction matters because achieving accurate bone preparation and implant positioning may theoretically have different implications when initial implant fixation depends on bone-implant contact and subsequent biological ingrowth.
It remains a separate question requiring appropriate study rather than extrapolation from RACER.
What Did RACER Tell Us About Cost?
The economics are equally important.
RACER concluded that robotic TKR, as delivered within the trial, was more costly than conventional TKR without producing a clinically meaningful patient benefit at 12 months.
The original protocol was specifically designed to evaluate cost-effectiveness alongside clinical outcomes. The study was funded primarily through the UK’s NIHR Health Technology Assessment program. Stryker provided specified treatment-related support, including consumables, preoperative CT costs and some operating-room time, under agreements intended to preserve the investigators’ independence in study design, data collection, analysis and interpretation.
Cost-effectiveness also depends heavily on context. Previous modeling has suggested that robotic TKA economics may change substantially with institutional volume, while other analyses have reached differing conclusions depending on assumptions regarding equipment costs, revisions and downstream healthcare utilization.
That makes the forthcoming detailed RACER health-economic work particularly important.
RACER Is Not Finished
Perhaps the most important point is that the current publication is not the final chapter.
The original RACER protocol includes follow-up at 2, 5 and 10 years, in addition to shorter-term endpoints.
Metcalfe told us that two-year data collection is complete and analysis is underway. Additional planned work includes health economics, surgeon learning effects and deeper analysis of the relationship between preoperative anatomy, intraoperative alignment decisions and postoperative outcomes.
Long-term implant survivorship may ultimately be especially important. A technology that produces no meaningful difference in FJS at one year could theoretically still have value if greater precision translates into fewer revisions or other downstream benefits. Conversely, if those benefits never emerge, the economic argument for widespread adoption becomes more difficult.
Those questions remain unanswered.
So, Does RACER Prove That Robotic Knee Replacement Doesn’t Work?
No. But it does challenge an overly simplistic argument for robotics.
RACER provides strong evidence that introducing robotic assistance alone does not automatically translate into better patient-reported outcomes at 12 months.
That finding deserves to be taken seriously.
At the same time, the study raises a potentially more interesting question for the next generation of research. If robotics gives surgeons increasingly reproducible control over implant positioning and soft-tissue balance, perhaps the limiting factor is no longer simply our ability to execute a plan.
Perhaps it is determining what the optimal plan should be for each individual patient.
That means the next phase of robotic knee replacement research may need to move beyond asking:
Can the robot put the knee where we tell it to?
The more important question may be:
Where should we tell it to put the knee?
Professor Metcalfe and Professor Davis discuss that question, the criticisms of RACER, its limitations, its funding, surgeon experience, alignment, costs and the future RACER studies in the full episode.
Watch Episode 41 of The Joint Replacement Podcast on YouTube
This content is intended for education and discussion and does not constitute individualized medical advice.