Metabolic endoscopy is increasingly recognised as a potential adjunct to conventional treatment for patients with type 2 diabetes (T2D). Current ablation modalities are limited by technical intricacies because of their over-the-guidewire design, which could limit real-world scalability. These systems used for duodenal mucosal ablation have been associated with prolonged procedure times relative to the length of tissue treated. An advancement of current ablation methods is proximal intestinal mucosal ablation (PIMA), an investigational endoscopic procedure for the treatment of T2D. The procedure aims to reversibly destroy up to 60 cm of the abnormal, hypertrophied proximal intestines with a through-the-scope Radiofrequency Vapor Ablation (RFVA) catheter. While ablating a longer segment of bowel may enhance glycaemic effect and durability, procedure skill acquisition is unclear. Therefore, we aimed to evaluate the feasibility and scalability of PIMA through a procedural learning curve. 

We evaluated procedure times among 39 patients undergoing PIMA for the treatment of T2D in a series of first-in-human studies (NCT06655740, NCT06724822, NCT05887635). All procedures were performed between Sep-2024 and May-2025 in Santiago, Chile, by two therapeutic endoscopists in five discrete treatment windows. Patients underwent PIMA via an adult colonoscope (Olympus HQ190L) with the through-the-scope RFVA catheter. Sequential ablations were delivered with double application to the post-ampullary intestinal mucosa with minimal overlap. Total procedure time was defined as the duration (minutes) from endoscope insertion to final removal. Catheter time was defined as the duration from RFVA catheter insertion to final removal. Time per ablation was calculated as total catheter time divided by the number of ablations delivered. Learning curves were modelled using non-linear regression of the form time = a + b/case number. Predicted values and 95% confidence intervals were generated to visualise the model fit. The first derivative of the fitted curve was used to identify the procedural plateau, defined as a reduction of <0.5 minutes per additional case (<0.05 minutes for time per ablation), beyond which further experience did not meaningfully reduce procedure time. Mean procedure duration at this plateau was considered the stabilised procedural time (procedural efficiency). Linear regression evaluated temporal changes in ablation number.

Across the cohort, average age was 50.2 years (SD 7.9) and 56.4 % were female. The mean procedure time was 62 minutes (SD 16.2), catheter time was 53 minutes (SD 14.2), and time per ablation was 1.14 minutes. The median ablated length was 60 cm (IQR 50-65). The learning curve for catheter time showed a plateau after approximately five cases, with a stabilised mean duration of 53.1 minutes. However, the non-linear learning-curve explained little of the variation in catheter time (R² = 0.01; p=0.535), which indicates no measurable reduction in catheter duration with increasing operator experience. A similar pattern was observed for total procedure time (R² = 0.01; p=0.483). The paradoxical initial increase in catheter and procedure times on the learning curve is driven by the increasing number of total ablations delivered, which was confirmed on linear regression (β = 0.40; R² = 0.21; p=0.004). When assessing true procedural efficiency, measured using time required to deliver each ablation, we observed a stabilised performance at 1.13 minutes per ablation after only five cases, with no further evidence of learning beyond this point.

PIMA is associated with a minimal ongoing learning effect, with stable time per ablation after five cases. Case-to-case variability appears driven by the complexity of enteroscopy with increasing treatment lengths rather than by catheter manipulation.