Robotics/Germany, Italy
845 patients, better on everything, and the scheduling system decided who got the robot
A quasi-randomised study of 845 knee replacements found robotic assistance better on every measure, with allocation by scheduling system.
A German and Italian team compared one-year patient-reported outcomes between robotic-assisted and conventional total knee arthroplasty in 845 patients at a single hospital. Allocation was by surgeon assignment through the scheduling system. The robotic cohort was significantly better across WOMAC, pain, activity and satisfaction measures, all p < 0.001.
What happened
A team at the University Hospital of Halle, Martin-Luther University Halle-Wittenberg, the Academic Hospital of Bolzano and Link Campus University published in the Journal of Orthopaedics and Traumatology a comparison of robotic-assisted and conventional total knee arthroplasty.
845 patients were included, operated at the Eifelklinik St. Brigida in Simmerath, Germany, between 2022 and 2025: 390 robotic-assisted and 455 conventional. At one year, 383 and 445 respectively completed follow-up.
The authors call it a quasi-randomised controlled trial, and are explicit about why: group allocation was determined by surgeon assignment through the institutional digital scheduling system.
At one year the robotic cohort showed significantly better results across WOMAC and its subscales, visual analogue scale pain, Tegner activity score, and functional and satisfaction items including walking distance and knee stability, all at p < 0.001.
The test
Two features of the design govern how the result should be read, and the paper states both.
The first is the allocation. Patients were assigned by which surgeon’s list they landed on, via the scheduling system. That is not randomisation. Whoever is available on a given day, and which patients get booked onto which list, can correlate with age, comorbidity, deformity severity and urgency. None of those are reported as balanced in the record abstract.
The second is more consequential. The two arms followed structurally different surgical workflows. The robotic arm used a four-dimensional personalised alignment protocol, covering coronal, sagittal, transversal and balancing planes, guided intraoperatively by a knee balancing score. The conventional arm used a femur-first technique with adjusted mechanical alignment.
So the comparison bundles the robot with a different alignment philosophy. Both patients received the same implant, a posterior-stabilised Smith and Nephew Legion Genesis II, which removes one confounder and leaves the larger one in place.
The numbers
| Item | Robotic-assisted | Conventional |
|---|---|---|
| Patients included | 390 | 455 |
| Completed one-year follow-up | 383 | 445 |
| Allocation | surgeon assignment via digital scheduling system | |
| Alignment philosophy | 4D personalised, Knee Balancing Score guided | femur-first, adjusted mechanical |
| Robotic system | CORI (Smith and Nephew) | – |
| Implant | Smith and Nephew Legion Genesis II, posterior-stabilised (both arms) | |
| WOMAC and subscales at 1 year | significantly better robotic, p < 0.001 | |
| VAS, Tegner, function, satisfaction | significantly better robotic, p < 0.001 | |
| Setting | single hospital, Simmerath, Germany, 2022 to 2025 | |
The result
Significantly better on every measure at p < 0.001, in 845 patients, is a striking set of numbers. It is also the pattern that should trigger the most scrutiny, for three reasons that compound.
Allocation was not random. The scheduling system decided, which means any factor correlated with list assignment is a potential confounder, and none is reported as controlled.
The arms differed in more than the robot. Comparing a personalised four-plane alignment workflow against a femur-first mechanical technique is comparing two surgical philosophies. The robot is one component of one of them.
Patients were not blinded. Every outcome reported is patient-reported, and patients knew whether a robot had been used. In a hospital that has invested in robotic arthroplasty, patients on the robotic list know they received the newer thing.
Uniform superiority across all outcomes, in an unblinded non-randomised comparison with differing workflows, is more consistent with those three features than with a large causal effect of robotic assistance.
How this compares to what came before
This is the comparison that matters and it is unflattering.
Pooled randomised evidence on robotic knee replacement has repeatedly found no patient-reported benefit. The accumulated randomised trials, pooled, find better implant positioning and no difference in what patients report.
Within this batch, the pattern across the robotic surgery cluster is the same inverse relationship between design strength and effect size.
The telesurgery trial, randomised and multicentre with a pre-specified margin and openly deposited data, found a difference of 0.02 in probability of surgical success. The colorectal trial, randomised and multicentre with 314 patients, used a biochemical endpoint. The liver resection post hoc analysis, a subgroup with imbalanced arms, found a large difference at p = 0.045. This study, with allocation by scheduling system and unblinded patient-reported outcomes, found significance on everything at p < 0.001.
Design strength runs one way and effect size runs the other. That is the most important thing in this batch’s surgical cluster, and this paper is the clearest illustration of its bottom end.
What it means
For orthopaedic practice, nothing here should change a decision about buying a surgical robot. The pooled randomised evidence remains the better guide.
What the paper may genuinely show is something different from what it claims: that a personalised four-plane alignment workflow with intraoperative balancing produces better one-year patient-reported outcomes than a femur-first mechanical technique. That is a real and interesting hypothesis, and it is testable without a robot. The robot is how this team implements the workflow, not necessarily why it works.
For readers, the transferable habit is to check the allocation mechanism before the p-values. The word quasi-randomised is doing a great deal of work in this paper, and the authors deserve credit for using it plainly rather than hiding it.
Business ideas from this paper
Alignment philosophy as the product rather than the hardware: if personalised four-plane balancing is what drives outcomes, it can be taught, protocolised and audited without capital equipment.
Intraoperative balancing measurement as a standalone tool, decoupled from a full robotic platform.
Registry-based outcome comparison for arthroplasty workflows, where large numbers can be adjusted for case mix in ways a single-centre series cannot.
What could be wrong with it
Allocation was by surgeon assignment through a scheduling system, which is not randomisation and leaves every list-correlated factor as a potential confounder.
The two arms followed different surgical workflows and alignment philosophies, so the robot is not the only variable.
All outcomes are patient-reported and patients were unblinded.
Significance on every measure at p < 0.001 in a non-randomised unblinded comparison is a pattern that invites caution rather than confidence.
Single hospital, single surgeon group, single implant, which aids internal consistency and limits generalisation.
Data are available on reasonable request rather than deposited.
845 patients is a large series, and size does not correct allocation bias; it narrows the confidence interval around a biased estimate.
How sure can you be?
Confident that 845 patients were operated at one hospital and that one-year patient-reported outcomes favoured the robotic cohort across the measures listed.
Confident that allocation was by surgeon assignment through the scheduling system, because the authors say so.
Confident that the two arms followed different alignment philosophies, which the paper describes.
Less confident that the difference is caused by robotic assistance rather than by the alignment workflow, the allocation mechanism or patient expectation.
Less confident still given that pooled randomised evidence has found no patient-reported benefit.
Not confident about generalisation beyond this hospital and this workflow pairing.
Our rubric places this at the floor of the range for this batch: peer-reviewed, four institutions, a prospective comparative design without randomisation, and data available on request.
Do this today
Check the allocation mechanism before the p-values. Here it was the hospital’s scheduling software, and that single fact outweighs 845 patients.
If you are comparing surgical technologies, ask what else differed between the arms. In this study, the alignment philosophy did, and it may be the active ingredient.
When a new technology is better on every measure at p < 0.001, look for the blinding. There was none here, and all the outcomes were patient-reported.
Source: Robotic-assisted total knee arthroplasty demonstrates superior early functional outcomes compared with conventional total knee arthroplasty: a quasi-randomised controlled trial, Journal of Orthopaedics and Traumatology, 2026-11-03. PMID 42980682 · doi.org ().
Just Out Tech explains new research in plain language. This article was drafted with AI assistance and checked by a human against the original source.
- 845 patients at the Eifelklinik St. Brigida in Simmerath, Germany, between 2022 and 2025: 390 robotic-assisted, 455 conventional.
- Group allocation was determined by surgeon assignment through the institutional digital scheduling system, which the authors describe as quasi-randomised.
- The two arms followed structurally different workflows: a four-dimensional personalised alignment protocol guided by a knee balancing score, versus femur-first adjusted mechanical alignment.
- At one year the robotic cohort was significantly better across WOMAC and subscales, visual analogue scale, Tegner activity score, and functional and satisfaction items, all p < 0.001.
Large single-institution comparison with allocation by surgeon scheduling rather than randomisation, and confounded by differing surgical workflows; unblinded patient-reported outcomes.
Questions people ask
Is quasi-randomised the same as randomised?
No. Allocation by surgeon through a scheduling system means whoever was available, or whichever list a patient fitted, determined their treatment. Those factors can correlate with patient characteristics.
What else differed between the arms?
The surgical philosophy. One arm used a four-dimensional personalised alignment workflow with intraoperative balancing; the other used a femur-first adjusted mechanical alignment technique.
Why does that matter?
Because the comparison is not robot against no robot. It is one alignment philosophy against another, with the robot bundled into one of them.
How does this sit with the pooled randomised evidence?
Pooled randomised evidence has repeatedly found no patient-reported benefit from robotic knee replacement. A single large non-randomised study finding benefit on every measure does not overturn that.