Pseudo-publication bias in robotic surgery: structural mapping of narrative inertia.

چکیده اصلی

The rapid worldwide adoption of robotic surgery often outpaces high-level comparative evidence. While evidence-based medicine relies on a hierarchical pyramid where causal inference resides at the apex, most of the scientific output is concentrated at the non-comparative base. It remains unknown whether the directional conclusions of these literature layers are symmetrically aligned. This study aimed to map the structural distribution of conclusions across the evidence hierarchy in visceral robotic surgery to evaluate potential narrative discordance. A stratified random sampling of intracavitary robotic surgery publications (thoracic, abdominal, and pelvic) was conducted from PubMed (1997-2026). Our sample size was calculated to achieve maximum representativeness (n = 800). Publications were equally partitioned into two pragmatic functional layers based on the presence of a control group: the Upper Hierarchy Zone (comparative layer, n = 400) and the Lower Hierarchy Zone (non-comparative layer, (n = 400). Directional orientations of author conclusions were operationally categorized as Favorable, Neutral, or Unfavorable. To ensure non-biased processing and scalability, semantic classification was executed via a Large Language Model (GPT-5.4), previously validated against a three-evaluator human pilot trial (absolute agreement 92%, Cohen's kappa > 0.81). Statistical analysis utilized Chi-square tests and Odds Ratio (OR) calculation with 95% Confidence Intervals (CI), with a secondary bipartite analysis (favorable vs. non-favorable). The global overview of the entire dataset (n = 800) demonstrated an illusion of literary equilibrium: 51.75% (n = 414) favorable, 45.50% (n = 364) neutral, and 2.75% (n = 22) unfavorable conclusions. However, stratification revealed a massive structural asymmetry (chi^2 = 312.9, p < 0.001). Within the upper hierarchy, conclusions were overwhelmingly non-favorable (79.50% vs. 20.50% favorable), dominated by neutral comparative outcomes (77.75%). Conversely, the lower hierarchy was heavily skewed toward favorable outcomes (83.00% favorable vs. 17.00% non-favorable). The lower zone also served as a sharper sensor of surgical failure, reporting a higher raw unfavorable rate (3.75%) than the apex (1.75%). Pragmatic bipartite analysis demonstrated that a publication at the base of the evidence pyramid has nearly 19 times higher odds of reporting a favorable conclusion compared to those at the comparative apex (OR 18.93, 95% CI: 13.84-26.87). A profound structural asymmetry exists within the robotic surgery literature. The disproportionate volume of highly favorable, lower-tier evidence numerically overwhelms comparative data, creating a phenomenon of "pseudo-publication bias." This structural distortion generates a powerful narrative inertia that drives global clinical adoption and shapes professional perception through cumulative enthusiasm rather than demonstrated methodological superiority.

متن کامل اصلی