Diagnostic Bronchoscopy for Peripheral Lung Lesions: When Innovation Races to Meet Demand

Transbronchial lung biopsy for peripheral pulmonary lesions is an airway procedure in which small tissue samples are taken from the lungs via endoscopic guidance in a targeted fashion. Since its advent in 1953 (1), the success of this methodology has been challenged by: 1) the progressively small size of identified lesions, 2) the increasing prevalence of peripheral lesions, 3) the lack of access to multidimensional intraprocedural imaging, and 4) poor and operator-limited bronchoscopic guidance techniques. The need for improved diagnostic approaches has intensified as more pulmonary nodules are discovered, in large part due to the emergence of lung cancer screening with low-dose computed tomography (CT) (2). In 2021, the U.S. Preventive Services Task Force broadened its lung cancer screening guidelines, further expanding eligibility to more than 13.5 million people (3, 4). This, in addition to the rapid increase in incidentally discovered pulmonary nodules (5), has created an imperative to accelerate the development of bronchoscopic techniques that provide improved access to the lung periphery.

Front and center among recent developments are electromagnetic navigational bronchoscopy (ENB) and robotic-assisted bronchoscopy (RAB). The former employs a magnetic field and sensors to locate the bronchoscope tip relative to a CT-generated three-dimensional image, allowing for more precise lesion targeting compared with conventional bronchoscopy. RAB platforms use electromagnetic guidance or shape-sensing technology to advance robotic catheters through the airway tree to effectively navigate to and sample peripheral lesions (6, 7). Both have been approved by the 510(k) pathway of the U.S. Food and Drug Administration, which allows introduction of a device into commercial distribution while only requiring “substantial equivalence” of safety and efficacy compared with prior technology (8). Hence, these systems were introduced into practice with an absence of robust comparative clinical outcome data.

In this issue of the Journal, Paez and colleagues (pp. 1644–1651) report on the RELIANT trial (Robotic versus Electromagnetic Bronchoscopy for Peripheral Pulmonary Lesions: A Randomized Trial), a comparative efficacy trial of ENB and shape-sensing RAB (9). This single-center, unblinded, cluster-randomized noninferiority trial compared diagnostic yield and safety in 411 patients with pulmonary lesions requiring bronchoscopic biopsy. Study participants were randomized on the morning of their procedure to undergo ENB or RAB. Careful attention was paid to the definition of diagnostic yield, whereby only the acquisition of specific lesional tissue explaining the presence of the pulmonary lesion was considered diagnostic (10). The importance of this cannot be overstated, as the literature on diagnostic bronchoscopy remains fraught with variable and inconsistent approaches to the reporting of accuracy, limiting our ability to derive meaningful conclusions from much of the prior literature.

Although mechanical ventilation parameters were standardized, the choice of biopsy tools and use of cone-beam CT were left to the discretion of the bronchoscopist. The diagnostic yield of RAB was noninferior to that of ENB (77.8% vs. 75.5%), and no difference was noted in the superiority analysis. The trial results largely confirmed a prior retrospective analysis from the same center, which reported yields of 80% and 77% for RAB and ENB, respectively (11). Although the procedure duration was longer in the RAB arm (37 min vs. 32 min), safety outcomes, including pneumothorax, respiratory failure, and hemorrhage, were infrequent and did not differ between study arms.

To our knowledge, this is the first cluster-randomized controlled trial directly comparing outcomes between ENB and RAB. Several aspects of this study design may provide distinct advantages. Trials that randomize individual patients are often resource-intensive and costly, and cluster-randomized designs can provide greater resource efficiency and reduce infrastructural complexities. Given the rapid pace of technological advances in peripheral bronchoscopy, the ability to perform trials efficiently while maintaining scientific rigor is critical to advancing our understanding of the advantages and limitations of these new diagnostic platforms.

Despite the potential advantages of cluster randomization, there remain inherent limitations of this design, including a greater risk of bias and the need for close attention to confounding variables in the statistical analysis (12). The RELIANT investigators incorporated several key design elements (e.g., blinding until the day of the procedure, daily randomization of operating rooms, small cluster size, regression modeling) that likely minimized bias and confounding. Additionally, there are some critical issues for investigators and clinicians to consider as future trials are designed and performed. First, the systems under evaluation do not perform a biopsy independently. A bronchoscopy platform requires an operator to make several calculated decisions during the procedure: use of intraprocedural imaging, choice of biopsy tools, and level of aggressiveness of sampling. How the interplay between the operator and a bronchoscopy system across different clinical scenarios affects diagnostic yield is unknown and should be carefully considered as we determine how best to design and conduct clinical trials that will inform clinicians on decision-making during bronchoscopy.

Second, RELIANT was conducted in a single academic center with considerable expertise in advanced diagnostic bronchoscopy. Based on the trial findings, the performance of ENB and RAB appear to be essentially equivalent in their hands. The expertise of the operators could have resulted in a higher level of performance with ENB, which is generally considered to have a steeper learning curve than RAB. Thus, proceduralist experience and ability may profoundly limit the generalizability of these findings to bronchoscopists with various levels of training and experience. In future trials, the inclusion of multiple centers and operators with varying levels of expertise would allow for analyses to examine this source of heterogeneity and result in improved generalizability.

Finally, the trial did not mandate the use of intraoperative imaging or specific biopsy tools. The contribution of these cointerventions on the effectiveness of diagnostic bronchoscopy systems remains poorly understood given the lack of prior evidence. Recent retrospective analyses have focused on the use of intraoperative cone-beam CT with RAB, suggesting yields that exceed 90% (13, 14). Many in the field believe these studies have already established the value of cone-beam CT given the known challenges of CT to body divergence and how intraprocedural atelectasis and temporal variance make navigation alone fraught with error, absent the ability to correct intraoperatively (15). Future trials will need to consider how best to incorporate adjunctive interventions, particularly the use of intraprocedural imaging, to facilitate robust effectiveness data that can inform future decision-making in real-world settings.

The trial design and approach used in RELIANT serve as an important framework in our approach to the evaluation of diagnostic bronchoscopy and may serve as a viable road map for future novel trials for a field yearning for tangible, high-quality evidence. Additional trials are needed to further advance our understanding of these emerging platforms and guide optimal practices in terms of safety and efficacy. We hope this trial will serve as a launching point to a new era of research in diagnostic bronchoscopy.