Robotic-Assisted Bronchoscopy for the Diagnosis of Lung Lesions: Experience With the Use of Frozen Sections as an Aid to Confirm the Localization of Lesions During the Procedure Open Access

Our institution introduced the use of R-ANB using the ION endoluminal system (Intuitive Surgical) in 2020. Specimens are brought by a laboratory assistant from the operating room to a central intraoperative consultation center located on another floor in the pathology department. Rapid onsite examination of cytologic specimens and frozen sections are prepared at this location by pathology residents and/or pathology assistants. The slides are interpreted by attending cytopathologists, surgical pathologists, and their residents working together. Turnaround times are recorded to ensure that they comply with College of American Pathologists guidelines. The results of most cases are conveyed within 20 minutes of specimen acquisition.

All consecutive intraoperative consultations with frozen section and transbronchial biopsies collected with the ION endoluminal system at Cedars-Sinai Medical Center, Los Angeles, California, during a 30-month period were reviewed. The patient’s age and gender and diagnostic information about their lung lesions were retrieved from the records of the Department of Pathology and Laboratory Medicine of Cedars-Sinai Medical Center and the hospital’s electronic medical records, with institutional review board (IRB 2217) approval; the study was performed in accordance with the Declaration of Helsinki. The biopsies obtained by R-ANB were obtained with 2 or more passes and included up to 7 specimens per procedure and up to 15 tissue fragments per specimen. During intraoperative consultations all tissue fragments were cut at 5-μm sections and stained with hematoxylin-eosin using an automated slide stainer. Residual tissue in the block used for frozen section and in all specimens obtained during the R-ANB procedure were fixed in formalin and processed into paraffin blocks. The paraffin blocks were cut in 4-μm sections, and all slides were stained with hematoxylin-eosin. It is difficult to cut frozen sections of multiple small tissue fragments, but in most cases the quality of frozen section slides was equivalent to slides cut from control formalin-fixed, paraffin-embedded tissue (Figure 1, A through D).

Pathologic diagnoses rendered during intraoperative consultations were labeled as frozen section diagnoses (FSDs). Pathologic diagnoses rendered on formalin-fixed tissue from specimens used for frozen section and/or additional specimens obtained during the same robotic assisted bronchoscopy procedure were labeled as formalin-fixed tissue diagnoses (FFTDs).

Diagnoses were classified into the following semiarbitrary classes: benign, adenocarcinoma with lepidic pattern (possible adenocarcinoma in situ, atypical adenomatous hyperplasia), small cell carcinoma neuroendocrine tumor/neuroendocrine carcinoma, granuloma/pneumonia, nonepithelial tumors (including melanoma, lymphoma, solitary fibrous tumor, and other metastatic lesions), atypical cells cannot exclude malignancy, and insufficient lung tissue for diagnosis. These classes were selected semiarbitrarily because the number of descriptive and/or specific diagnoses included too many categories to evaluate individual diagnostic classes without a much larger number of cases.

The concordance percentages between FSD and corresponding FFTD were calculated by diagnostic category. The sensitivities and specificities of FSD and FFTD were calculated by diagnostic category using FFTD and/or FD as true positives. In addition, the percentage of biopsies with lung adenocarcinoma that included sufficient tissue for the performance of next-generation sequencing, fluorescence in situ hybridization for echinoderm microtubule associated protein like 4::anaplastic lymphoma kinase (EML4::ALK) fusion, MET proto-oncogene, receptor tyrosine kinase (MET), and ROS proto-oncogene 1, receptor tyrosine kinase (ROS1) and immunohistochemistry for programmed death-ligand 1 (PD-L1) was calculated.

Our cohort consisted of 203 patients with 383 lesions. Thoracic surgeons sent 226 of 383 (59%) lesions for intraoperative consultation with frozen section during R-ANB during a 30-month period. The remaining 157 lesions were biopsied by pulmonologists who did not request frozen sections. As pulmonologists do not perform single-anesthesia surgery and most biopsies can be reported in our department within 24 hours, they did not consider that frozen sections were necessary. One hundred and seven of the 203 (53%) patients had malignancies, as confirmed with FFTD, wedge biopsies performed in patients with negative R-ANB, and/or lung resection. Malignancies were diagnosed intraoperatively with frozen section with 100% specificity but only modest sensitivity of 68% using the following categories (Figure 2, A through F): adenocarcinoma with lepidic pattern (n = 4), non–small cell lung carcinoma (n = 53), neuroendocrine tumor/carcinoma (n = 7), nonepithelial tumors (including melanoma/lymphoma/solitary fibrous tumor/metastases) (n = 10), and atypical cells (n = 33) (Table). The atypical cells category comprised cases where the pathologist could not rule out malignancy due to limited specimen during the frozen section or the paraffin-fixed tissue diagnosis. Ninety-one of the 203 (45%) patients had nonneoplastic conditions that were diagnosed intraoperatively as benign (n = 76), or granuloma/pneumonia (n = 15). The remaining cases (n = 8) rendered insufficient tissue for diagnosis.

Molecular testing was requested on biopsies from 52 lung adenocarcinomas, and 48 (92%) of these biopsies had sufficient residual tissue after frozen section for the performance of next-generation sequencing, immunohistochemistry for PD-L1, and fluorescence in situ hybridization for EML4::ALK fusion, MET, and ROS1. Four biopsies from the 107 patients with malignancies had insufficient tissue for molecular studies after the performance of frozen section. These 4 patients and 103 others with malignancies that were not diagnosed by frozen section because the test was not ordered or the results were negative after the initial robotic bronchoscopy biopsies had to undergo a subsequent procedure to obtain sufficient tissue for diagnosis and/or molecular testing because the initial biopsies were inadequate.

Our results show that intraoperative consultations with frozen section provide an effective tool to confirm that peripheral lung lesions have been adequately targeted and biopsied by R-ANB and to diagnose a variety of neoplastic and granulomatous lesions with high specificity and modest sensitivity. They were performed in 226 instances during the workup of 383 lesions in cases when the interventional pulmonologist or thoracic surgeon needed to confirm that the lesion was adequately targeted or the patient could undergo surgery during a single anesthesia. Biopsies from the remaining 157 lesions were submitted in formalin for processing and diagnosis within 24–48 hours. FSDs were reasonably accurate, with 72% overall concordance with FFTDs, and malignancies were diagnosed with 100% specificity.

The development of an interdisciplinary R-ANB program is costly for pathology departments, as the performance of frozen sections and concomitant cytologic interpretations on multiple specimens from a single patient can be very challenging and time consuming and requires adequate staffing by pathologists who are experienced with the interpretation of frozen sections and the rapid evaluation of cytology smears and are comfortable diagnosing lung neoplasms using very small samples. The program in our hospital has not increased costs, as our intraoperative consultation service is routinely staffed by a surgical pathologist, a pathology resident, and a cytopathologist who is summoned only when needed. These staffing levels, necessary to provide intraoperative consultations to surgeons from multiple specialties, have not been increased because of this new program. Moreover, the costs of performing frozen sections and rapid onsite evaluation of cytologies from patients with pulmonary lesions look favorable from the patient’s and institution’s viewpoints. Indeed, intraoperative diagnoses rendered during R-ANB at Cedars-Sinai Medical Center are very helpful to our thoracic surgeons, as this allows for the selection of patients who can undergo lobectomy or other lung resections using single anesthesia the same day of bronchoscopy, and/or the scheduling of individuals for surgery the day after bronchoscopy. Intraoperative diagnosis can also be helpful to our lung pathologists to streamline the diagnostic process, as they can use this information to promptly order histochemical stains, immunohistochemical stains, and/or molecular studies, contributing to the shortening of hospital stays and the streamlining of therapeutic processes.

An International Association for the Study of Lung Cancer survey in 2020 reported that the main reason for failure in molecular studies is insufficient tumor cells and inadequate tissue quality.⁶  A concern when R-ANB was introduced to our patients was that the small biopsy samples needed very careful handling during the processing of frozen section slides. For example, tissue fragments could be lost, sections could fall off slides during staining, and small tissue samples could be completely cut while performing frozen section, impairing the subsequent performance of molecular studies and subjecting patients with lung carcinomas to additional invasive procedures to obtain adequate tissue samples. Unfortunately, we did have a few initial cases where the surgical pathologist was so anxious to provide an accurate diagnosis that too many frozen section slides were cut, leaving insufficient residual tissue for molecular studies. We also had several cases where the biopsies consisted mostly of red cells and had only scant fragments of diagnostic tissue that were too small for subsequent molecular studies. Our surgical pathologists now recognize the need to balance diagnostic and tissue preservation needs and routinely request additional tissue for molecular studies in cases that are likely to yield insufficient materials for next-generation sequencing or other ancillary studies. Thoracic surgeons and pulmonologists also learned to submit biopsies fixed in formalin, in addition to tissue that needs to be examined intraoperatively. As a result of our combined efforts, only 4 out of 52 (8%) biopsies from lung cancer patients had insufficient tissue for molecular studies after frozen section, and most of the inadequate cases were performed during the initial learning curve. As there is a learning curve to optimize the clinical value of R-ANB, pathologists participating in the development of these interdisciplinary programs need to be careful to approach the interpretation of frozen sections from these cases with caution to avoid significant diagnostic errors and/or medicolegal complaints.

There is still limited experience in literature with the use of R-ANB for the diagnosis of peripheral lung lesions. The PRECISE Study of the ION endoluminal system, a multistage, single-arm, prospective evaluation of the ION endoluminal system, reported that navigators were able to target small peripheral pulmonary nodules safely; however, the diagnostic yield and sensitivity associated with each nodule warranted further investigation.⁴  Other studies have reported a diagnostic yield of 83% while using ION-navigated robotic-assisted bronchoscopy with lesion localization of 97%.^6,7  To our knowledge, there are no previous studies reporting the logistic and other challenges that R-ANB presents to pathologists and their clinical colleagues. A learning curve is expected, both for bronchoscopists to obtain adequate samples and for pathologists who need to render diagnoses on minute tissue fragments measuring 1 to 3 mm each. Laboratories need to be aware that multiple specimens are often sent simultaneously to the frozen section laboratories during R-ANB, creating logistics and staffing challenges.