This Week in Surgery — May 14, 2026

Welcome to This Week in Surgery. This week we're covering 10 notable papers spanning nuances in surgical oncology, outcomes in complex pancreatic surgery, and the evolving role of artificial intelligence in our field. We'll also touch on some foundational science with direct clinical relevance. Let's dive in.

Artificial Intelligence and Technology

First, we look at the integration of artificial intelligence and technology into surgical practice, where two new studies offer both promise and a dose of reality.

The Study

In the Annals of Surgical Oncology, investigators examined an AI-driven three-dimensional reconstruction system for preoperative planning in thoracic surgery [4]. They retrospectively compared over one thousand patients who underwent segmentectomy or lobectomy planned with either the AI-3D system or standard two-dimensional CT scans. The primary goal was to see if the advanced imaging reduced unexpected changes to the surgical plan.

Results

For patients scheduled for segmentectomy, the AI-3D system made a significant difference. The consistency between the planned and actual procedure was 97% in the AI group, compared to just 80% in the standard CT group. This suggests the technology provides a more reliable preoperative roadmap, primarily by better clarifying tumor location and potential surgical margins. However, this planning advantage did not translate into better intraoperative or postoperative outcomes. Operative time, blood loss, and complication rates were comparable between the two groups.

Discussion

While AI-driven reconstruction can enhance surgical reliability, another study in Surgery sounds a note of caution regarding AI in high-stakes clinical decision-making [2]. Researchers tested three leading large language models—including families from OpenAI, Google, and Anthropic—on their ability to perform trauma triage using 150 standardized vignettes. The primary concern was undertriage, or incorrectly classifying a major trauma patient as having minor injuries.

Results

Overall, the models performed well, with a combined sensitivity for major trauma of nearly 90%. However, the errors were not random. The models were significantly more likely to undertriage geriatric patients and those with blunt trauma compared to younger adults or those with penetrating injuries. These are the same patient groups that are notoriously challenging for human clinicians. Furthermore, the study revealed what the authors call 'prompt fragility'—slight changes in the wording of the clinical scenario caused the AI to change its triage decision over 11% of the time. This instability, combined with the systematic bias against vulnerable populations, suggests these models should be used, if at all, as supervised adjuncts rather than autonomous triage agents.

Nuances in Surgical Oncology

Next, we turn to surgical oncology, where three papers challenge a one-size-fits-all approach and instead advocate for more tailored, and sometimes less aggressive, treatment strategies.

The Study

First, a major analysis from the National Cancer Database, published in Surgery, questions the routine use of neoadjuvant chemotherapy in a large subgroup of breast cancer patients [5]. The study included over 146,000 patients with node-positive, hormone receptor-positive, HER2-negative breast cancer. Investigators compared overall survival between patients who received neoadjuvant chemotherapy before surgery and those who had upfront surgery.

Results

The results were striking. After adjusting for baseline differences, neoadjuvant chemotherapy was associated with inferior overall survival compared to upfront surgery. The survival disadvantage was most pronounced in patients with a lower nodal burden, or N1 disease. The one group that benefited significantly were the 10% of patients who achieved a pathologic complete response in their lymph nodes after chemotherapy; they had excellent long-term survival. The authors conclude that careful patient selection is paramount, as the benefits of neoadjuvant therapy in this population appear to be limited to a small, chemosensitive subgroup.

Discussion

This finding dovetails with a study in the Annals of Surgical Oncology that explores the adoption of another de-escalation strategy in breast cancer: the omission of sentinel lymph node biopsy based on the SOUND trial [3]. In a survey of North American physicians, researchers found that while over 60% reported a high intent to implement the SOUND trial's findings, actual practice change is slow. Factors associated with a greater intent to omit axillary surgery included a longer awareness of the trial's results and practicing in the Northeast United States. This highlights the persistent gap between the publication of practice-changing evidence and its widespread adoption.

Finally, a study in Surgery provides clarity on a common dilemma in thyroid cancer: the significance of a microscopic positive margin [1]. In a cohort of 502 patients with low-risk T1-T2 well-differentiated thyroid cancer, 13% had a positive microscopic margin after thyroidectomy.

Results

Microscopic margin positivity was independently associated with an increased risk of locoregional recurrence, nearly tripling the risk compared to patients with negative margins. However, and this is the crucial nuance, margin status had no impact on the likelihood of achieving an excellent response to therapy. The clinical implication is that while a positive microscopic margin should prompt more vigilant surveillance, it may not necessitate routine re-operation or other treatment escalation, as it doesn't appear to compromise the ultimate therapeutic response.

Minimally Invasive Pancreatic Surgery

Our third theme focuses on pancreatic surgery, with two studies examining outcomes and challenges in minimally invasive approaches.

The Study

A nationwide study from Norway, published in the Journal of the American College of Surgeons, used a national registry to evaluate the implementation of minimally invasive left-sided pancreatectomy [7]. The study analyzed 817 procedures performed between 2016 and 2023, of which about two-thirds were done laparoscopically.

Results

The investigators found that minimally invasive surgery was associated with a two-fold higher rate of achieving an 'ideal outcome,' which is a composite measure including shorter hospital stay and fewer complications. However, this came with a major caveat: significant selection bias. Patients selected for the laparoscopic approach were younger, healthier, and had a lower American Society for Anesthesiologists score. The study also revealed significant regional variation in the adoption of minimally invasive techniques, suggesting that access and practice patterns are not uniform.

Discussion

While that study looked at left-sided pancreatectomies, a large multi-center study from the European E-MIPS registry, published in Annals of Surgery, focused on a critical event in minimally invasive pancreatoduodenectomy: conversion to open surgery [8]. The study compared outcomes after conversion from either a robotic or a laparoscopic approach. After propensity score matching, conversion rates were similar between the two platforms at around 12%.

Results

The key finding was that conversion during a robotic pancreatoduodenectomy was associated with significantly worse outcomes. Major morbidity was higher in the robotic conversion group compared to the laparoscopic conversion group, at 49% versus 31%. This difference held true even in high-volume centers, where major morbidity was 48% for robotic conversions versus 29% for laparoscopic conversions. The authors suggest that the dynamics of converting from a robotic platform may be different and potentially riskier, warranting structured decision-making protocols and specific team training on robotic-to-open conversion.

Foundational Science for Clinical Practice

Finally, we'll briefly cover three papers from high-impact journals that provide foundational knowledge relevant to all practicing physicians.

The Study

First, a randomized clinical trial in JAMA investigated a potential treatment for osteogenesis imperfecta, a condition that causes multiple fractures [6]. Adults were randomized to receive either two years of the anabolic agent teriparatide followed by zoledronic acid, or standard care.

Results

This was a negative trial. The intensive drug regimen did not reduce the risk of fracture compared to standard care. This was despite the fact that the treatment group experienced significant increases in lumbar spine and hip bone mineral density. The finding strongly suggests that in osteogenesis imperfecta, poor bone quality, rather than low bone density, is the primary driver of fracture risk, which has major implications for future therapeutic targets.

Discussion

Next, two papers in Nature introduce novel 'charts' for human health. The first presents the first-ever lifespan reference charts for the brain's white matter [9]. By standardizing over 35,000 brain scans, researchers have created a normative benchmark for how the brain's communication pathways grow, mature, and decline from birth to 100 years of age. Much like pediatric growth charts, these brain charts will allow clinicians to quantify how an individual’s brain deviates from the norm, potentially aiding in the diagnosis and monitoring of numerous neurological and psychiatric disorders.

Our final paper, also in Nature, proposes a 'Sleep Chart' that links sleep duration to biological aging [10]. Analyzing data from the UK Biobank, the study found a consistent U-shaped relationship between self-reported sleep duration and biological age across nine different organ systems. The lowest biological age—meaning the healthiest state—was associated with sleep durations between 6.4 and 7.8 hours. Both short sleep, under 6 hours, and long sleep, over 8 hours, were associated with accelerated biological aging and increased risk for diseases like depression and diabetes.

Editor's Pick

If you only have time for one paper this week, make it the National Cancer Database analysis in Surgery on neoadjuvant chemotherapy for hormone receptor-positive breast cancer [5]. Its finding that neoadjuvant chemotherapy is associated with worse survival compared to upfront surgery in this group is a significant challenge to current practice and demands a careful reconsideration of patient selection for this common scenario.

Clinical Bottom Line

Here are the key takeaways from this week in Surgery.

First, in node-positive, hormone receptor-positive, HER2-negative breast cancer, neoadjuvant chemotherapy may be associated with worse survival than upfront surgery. Consider reserving it for patients you believe have a high chance of achieving a nodal complete response [5].

Second, when converting a minimally invasive pancreatoduodenectomy to an open procedure, be aware that conversion from a robotic approach may carry a higher risk of major morbidity than conversion from a laparoscopic approach, suggesting a need for specific protocols and training [8].

Third, AI tools require cautious optimism. AI-driven 3D reconstruction can improve surgical plan consistency in thoracic surgery [4], but large language models for trauma triage may systematically undertriage high-risk patients like the elderly and those with blunt trauma [2].

Fourth, for patients with low-risk differentiated thyroid cancer, a microscopic positive margin increases recurrence risk but does not appear to prevent an excellent response to therapy. This supports using margin status for surveillance planning rather than automatic treatment escalation [1].

And finally, for your patients with osteogenesis imperfecta, a major trial in JAMA found that teriparatide plus zoledronic acid did not reduce fracture risk, despite improving bone density, shifting the focus to bone quality [6].

That's your roundup for This Week in Surgery. The full transcript and references are available on the episode page in your AudioScholar library. This is an AI-curated summary — for clinical decisions, always consult primary sources and current guidelines. See you next week.

This is an automated summary generated by artificial intelligence, which can make mistakes. Always review the original source materials.