Abstract Background Robotic-assisted bronchoscopy (RAB) may improve diagnosis of peripheral pulmonary lesions (PPLs), particularly when lesion localization and sampling are technically demanding. However, direct real-world comparisons between shape-sensing RAB (ss-RAB) and virtual bronchoscopic navigation (VBN) remain limited. Methods We retrospectively reviewed patients who underwent navigational bronchoscopy for PPLs at a single center. Propensity score matching (PSM) was used to balance baseline characteristics between the ss-RAB and VBN groups. Diagnostic yield, tool-in-lesion (TIL) rate, sensitivity for malignancy, adverse events, and learning curves of ss-RAB were evaluated. Learning curves were assessed using cumulative sum analysis based on procedure time and diagnostic success. Results Following 1:1 PSM, 234 patients were included, with 117 in each group. Final malignancy prevalence was 74.8% overall. Diagnostic yield was higher with ss-RAB than with VBN (85.5% vs. 70.1%; OR 2.39, 95% CI 1.25–4.56). TIL rate was also more frequently achieved with ss-RAB (99.1% vs. 88.9%). Sensitivity for malignancy was 95.6% for ss-RAB compared with 81.2% for VBN. The diagnostic advantage of ss-RAB was most evident in small (< 15 mm), peripheral, bronchus-sign-negative, and upper-lobe lesions, whereas performance was similar in larger or more central nodules. Procedure-related complications were infrequent and comparable between groups. Learning-curve analysis for ss-RAB showed a reduction in procedure time after approximately 23 cases and an earlier improvement in diagnostic success around the 13th case. Conclusions In a real-world setting, ss-RAB was associated with higher diagnostic yield and TIL rates compared with the VBN-based comparator arm without compromising safety. The observed benefit of ss-RAB was greater in lesion subgroups with higher procedural difficulty. In experienced centers, procedural performance with ss-RAB improved after a relatively short learning phase.
You et al. (2026) studied this question.