When heterogeneity dominates: interpreting TBS-UAS benefits in RIRS with prediction intervals

Dear Editor, We read with great interest the systematic review and meta-analysis comparing tip-bendable suction ureteral access sheaths (TBS-UAS) with traditional sheaths in retrograde intrarenal surgery[1]. The authors are to be commended for addressing a rapidly evolving adjunct in retrograde intrarenal stone surgery (RIRS). From a forward-looking surgical and meta-analytic perspective, we would like to raise several methodological issues that were not discussed but may meaningfully influence the reliability and clinical interpretation of the pooled effects. Transparency In the reporting of Artificial Intelligence – The TITAN guideline[2]. First, the meta-analysis appears to treat all TBS-UAS systems as a homogeneous intervention. In practice, “tip-bendable suction” represents a family of devices and operating strategies rather than a single standardized technology. Across contemporary centers, suction may be continuous or intermittent, negative-pressure ranges differ, the suction port design varies, and some systems are integrated with pressure-controlled irrigation, while others are not. These technical parameters are not trivial: they directly determine fragment evacuation efficiency and intrarenal pressure modulation, the two mechanisms invoked to explain superior stone-free rates and lower infectious morbidity. Pooling nonequivalent TBS-UAS implementations risks averaging over devices with materially different performance, producing either diluted or inflated effects. At minimum, a structured description of suction mode, pressure range, and sheath tip controllability per study, followed by stratified or exploratory meta-regression analyses, would strengthen causal attribution to the device class rather than to high-performing subsets[3]. Second, the comparator arm is implicitly assumed to be “traditional UAS alone,” yet the clinical effectiveness of a standard sheath in RIRS is inseparable from co-interventions such as basket-assisted extraction frequency, endoscope withdrawal protocols for fragment egress, and the decision to pursue dusting versus active clearance. These practice patterns vary widely between surgeons and institutions and may systematically differ when a novel device is introduced. If control groups in some trials used more conservative fragment retrieval strategies, the observed advantage of TBS-UAS could partly reflect a lower “operation intensity” in the traditional-UAS arm rather than a true device effect. Future primary studies should report fragment-retrieval policies and basket use rates, and future syntheses should consider these as effect modifiers instead of assuming a neutral comparator[4–6]. Third, while reduced postoperative fever is a key finding, infectious outcomes in RIRS are highly confounded by baseline risk that is not uniformly controlled across studies. Preoperative urine culture positivity, the proportion of infection-related stones, and the exact antibiotic prophylaxis and “culture-clearance” protocols can outweigh the impact of intrarenal pressure alone. Novel devices are often adopted first in high-resource centers with stringent infection pathways, which may bias fever reduction in favor of TBS-UAS. Without balancing or adjusting for these baseline factors, attributing the magnitude of infectious benefit solely to suction-mediated pressure control is precarious[7]. We encourage explicit reporting of culture status and prophylaxis regimens, and sensitivity analyses stratified by preoperative infection risk where data permit. Finally, the pooled immediate stone-free outcome shows extremely high heterogeneity, yet only average risk ratios are presented. In the context of a new technology with variable implementation and surgeon familiarity, prediction intervals would be particularly informative to clinicians by conveying the plausible range of effects in future settings. Relatedly, device learning curves may meaningfully shape both efficacy and safety early after adoption. Integrating prediction intervals and exploring time-since-adoption or operator-experience proxies could prevent over-generalization of a mean estimate to all centers; in this regard, we provide an illustrative forest plot for the immediate stone-free outcome that adds a 95% prediction interval using the study-level event data and an analysis specification consistent with the original report (Fig. 1). Figure 1.: Prediction interval-augmented forest plot for immediate stone-free rate (SFR) comparing TBS-UAS versus traditional UAS in RIRS. We believe addressing these points would enhance the interpretability and translational value of this important work and guide more definitive multicenter trials. We appreciate the opportunity to contribute to the discussion.