Just out todayDigital health tech: Ninety-nine nurses, seventy wearable features, and no control groupDigital health tech: Fifty-six trials, 6,154 patients, and headsets beat the threshold patients would noticeSynthetic biology & cell therapy: Roughly one in four patients on these therapies gets a severe infection

Robotics/Germany

13% against 37% severe complications, from a post hoc subgroup with imbalanced arms

A post hoc analysis found severe complications in 13% after robotic and 37% after laparoscopic anatomical hepatectomy.

What the paper found

A post hoc analysis of the German ROC'N'ROLL randomised trial examined the 57 of 81 participants scheduled for anatomical hepatectomy. Severe complications occurred in 13% after robotic and 37% after laparoscopic resection, odds ratio 0.27 (95% CI 0.07 to 0.97, p = 0.045), despite larger tumours and higher difficulty scores in the robotic arm and longer operating times.

Grade 11 reading level6 min read

What happened

A German team at Ulm University Hospital, Universitatsmedizin Mannheim and Heidelberg University published in Surgical Endoscopy a post hoc analysis of the ROC’N’ROLL randomised controlled trial (DRKS00027531), comparing robotic and laparoscopic liver resection.

The analysis looks at the subset scheduled for anatomical hepatectomy, the version of the operation that follows the liver’s internal vascular divisions and is the technically hardest.

Of 81 randomised participants, 57 were scheduled for anatomical resection: 30 robotic, 27 laparoscopic.

Severe postoperative complications, Clavien-Dindo grade III or above within 90 days, occurred in 13% after robotic and 37% after laparoscopic resection. Odds ratio 0.27, 95% CI 0.07 to 0.97, p = 0.045.

The arms were not balanced. The robotic group had larger tumours, median 59 against 40 mm, p = 0.048, and higher difficulty scores, median IWATE 10 against 7, p = 0.029. Robotic operations took longer, median 252 against 190 minutes, p = 0.021.

The test

The clinical logic for robotic liver resection is good. Anatomical hepatectomy requires fine dissection along vascular planes in a solid organ that bleeds, and the robot’s advantages, wristed instruments and stable three-dimensional vision, should matter most there.

The methodological position of this analysis is weaker than the headline suggests, in a specific and important way. ROC’N’ROLL randomised 81 patients. This analysis takes the 57 who were scheduled for anatomical resection and compares them. Randomisation guarantees comparability across the whole trial population; it does not guarantee it within a subgroup defined after the fact, and here it demonstrably did not deliver it.

The endpoint itself is good: Clavien-Dindo grade III or above at 90 days is a counted, graded clinical outcome, not a rating.

The numbers

Measure Robotic (n=30) Laparoscopic (n=27)
Severe complications (CD ≥ III), 90 days 13% 37%
Odds ratio 0.27 (95% CI 0.07 to 0.97), p = 0.045
Median tumour size 59 mm 40 mm, p = 0.048
Median IWATE difficulty score 10 7, p = 0.029
Median operating time 252 min 190 min, p = 0.021
Parent trial ROC’N’ROLL, 81 randomised, DRKS00027531
Analysis type post hoc subgroup

The result

A reduction in severe complications from 37% to 13%, with an odds ratio of 0.27 whose confidence interval excludes one, is a large effect.

Three features should hold the reader back.

It is a post hoc subgroup of 30 against 27 patients. Subgroup analyses of randomised trials produce false positives at a well-documented rate, and a p-value of 0.045 from 57 patients is fragile.

The baseline imbalances cut in an awkward direction. Larger tumours and higher difficulty scores in the robotic arm make the better outcome more impressive if the result is real, and simultaneously demonstrate that the two groups were not comparable. A reader cannot have it both ways: if the arms differ enough to make the robotic result more impressive, they differ enough to confound it.

The 37% rate in the laparoscopic arm is high for anatomical liver resection. That raises a question the paper cannot answer: was the comparator performing to standard?

And robotic resection took 62 minutes longer, which is a real cost in theatre time.

How this compares to what came before

The robotic surgery evidence in this batch sorts cleanly by design strength, and this paper sits in the middle.

At the top, the telesurgery trial: randomised, multicentre, pre-specified non-inferiority margin, surgeon-clustered Bayesian analysis, data openly deposited at Dryad. At the next level, the colorectal trial: randomised, multicentre, 314 patients, biochemical endpoint. Then this: a post hoc subgroup of a randomised trial with a good clinical endpoint and imbalanced arms. At the bottom, the knee arthroplasty study, where allocation was by surgeon through a scheduling system and the robot was significantly better on every measure.

The ordering matters because the effect sizes run in the opposite direction. The most rigorous design found non-inferiority with a difference of 0.02. The least rigorous found superiority on everything at p < 0.001. This paper, in between, found a large difference in a subgroup with imbalanced arms.

That inverse relationship between design strength and effect size is the single most useful pattern in this batch’s surgical cluster, and it is the reason to read methods before results.

What it means

For liver surgery, this is a reason to run a trial powered for anatomical hepatectomy specifically, not a reason to change practice. The signal is interesting and the design cannot support it.

For reading subgroup analyses generally, the discipline is to ask three questions: was the subgroup pre-specified, were the arms balanced within it, and how many patients. Here the answers are no, no and 57.

For theatre planning, the 62-minute difference in operating time is the most solidly established finding and the one with an immediate operational consequence.

Business ideas from this paper

Difficulty-stratified trial design for hepatobiliary surgery, using IWATE scores to randomise within strata so that subgroup comparisons are not confounded by case mix.

Complication benchmarking services, since a 37% severe complication rate in a trial arm is the kind of number a unit would want to compare against its own.

Theatre scheduling models that price the robotic time penalty explicitly: 252 against 190 minutes changes how many cases fit on a list.

What could be wrong with it

It is a post hoc subgroup analysis of 57 patients, not the trial’s randomised comparison, so it does not carry the trial’s protection against confounding.

The arms were imbalanced on tumour size and difficulty score, both at conventional significance, which is evidence that randomisation did not balance this subgroup.

A p-value of 0.045 with an upper confidence bound of 0.97 is at the edge of conventional significance.

The 37% severe complication rate in the laparoscopic arm is high, and the paper cannot establish whether the comparator was performing as expected.

Operating time was 62 minutes longer with the robot, which offsets some of the claimed benefit.

No data availability statement appears in the indexed record.

How sure can you be?

Confident that the endpoint is a counted, graded clinical outcome at 90 days, which is the right kind of endpoint.

Confident that severe complications were less frequent in the robotic arm in this subgroup, and that operating time was longer.

Confident that the arms were imbalanced on tumour size and difficulty.

Less confident that the complication difference reflects the surgical approach rather than the imbalance, the subgroup selection or chance, given 57 patients and p = 0.045.

Less confident about the comparator’s performance, given a 37% severe complication rate.

Not confident that this would replicate in a trial designed to answer it.

Our rubric places this at the floor of the range for this batch: peer-reviewed, three institutions, a post hoc analysis of randomised trial data with a clinical endpoint, and no public data deposit.

Do this today

When you see a striking subgroup result, check whether the subgroup was pre-specified and whether its arms were balanced. Here neither was true.

If you perform liver resection, the operating time figure is the one to plan around: 252 against 190 minutes median.

If a trial arm reports an unusually high complication rate, ask how it compares with your own. A 37% severe complication rate is a finding in itself.

Source: Robotic versus laparoscopic anatomical hepatectomy: a post hoc analysis of the ROC'N'ROLL randomized controlled trial, Surgical Endoscopy, 2026-09-29. PMID 42811178 · 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.

What to remember
  • Post hoc analysis of the ROC'N'ROLL randomised trial: 57 of 81 participants were scheduled for anatomical hepatectomy, 30 robotic and 27 laparoscopic.
  • Severe complications (Clavien-Dindo grade III or above) within 90 days: 13% robotic versus 37% laparoscopic, odds ratio 0.27 (95% CI 0.07 to 0.97), p = 0.045.
  • Baseline imbalances favoured the laparoscopic arm: larger tumours in the robotic group, median 59 vs 40 mm, p = 0.048, and higher IWATE difficulty scores, median 10 vs 7, p = 0.029.
  • Operating time was longer with robotic resection, median 252 against 190 minutes, p = 0.021.
Where this really is

Post hoc subgroup analysis of a randomised controlled trial with a clinical complication endpoint; small subgroup with baseline imbalances.

Questions people ask

What is a post hoc subgroup analysis?

An analysis of part of a trial population, decided after the trial was designed. It does not inherit the randomised comparison's protection against confounding within that subgroup.

Why do the baseline imbalances cut both ways?

Larger tumours and harder cases in the robotic arm make its better result more impressive, and also show the arms were not comparable, which is exactly what randomisation is supposed to prevent.

Is 37% severe complications high?

Yes. For anatomical liver resection that is at the upper end of published rates, which raises the question of whether the comparator performed as it should.

What is the IWATE score?

A difficulty score for laparoscopic liver resection. A median of 10 against 7 means the robotic cases were substantially harder.

About the author

Mark Alex

Mark Alex is the founder and Managing Director of Real Biz Digital, a technology company operating out of Nairobi since 2018. He works in agentic AI and the Model Context Protocol, AI governance, enterprise software architecture and cybersecurity. He holds an MSc in Mechatronical Engineering from Obuda University in Budapest and a BSc in IT, Forensic Technology and Cybercrime, from USIU-Africa in Nairobi, and has published IEEE conference research on an AI-powered digital twin for greenhouse systems. He is the author of seven books. Between 2020 and 2024 he mentored more than 200 university students and interns in Nairobi. He writes every Just Out Tech article from the original research paper.