Because of the success of targeted cell therapies for treating non-small cell lung cancers, the need for accurate diagnostic capabilities is paramount. EBUS-TBNA is the standard, minimally invasive method to obtain diagnostic samples, and can be used to obtain material for next generation sequencing. Adequate collection of samples is essential to guide treatment decisions based on driver mutation status of ALK, EGFR, and ROS-1. However, there are no definitive guidelines on procedure protocols needed to obtain adequate samples for NGS testing.

In our study we analyzed samples from 278 patients with non-squamous non-small cell lung cancers, of which, 146 underwent NGS using tissue samples. When the protocol was adhered to, higher NGS yields were obtained, 66.0% versus 37.2% (p  < 0.01); however it was only followed 64.0% of the time. Despite lower than expected adherence to the protocol, we found that samples were sufficient for EGFR, ALK, and ROS-1 testing > 70% of the time. This number fell to 57.5% for NGS testing. We identified the only consistent predicting factor for sufficiency to be larger lesion size for EGFR, ALK, and ROS-1 testing, and number of aspirations for NGS testing. Reasons for differences in significant factors between NGS testing and individual mutation testing is unclear but may be due to laboratory processing requirements or the need to obtain more tissue to run multiple genetic tests with NGS rather than just one specific mutation. NGS testing was requested across all stages of lung cancer but was not found to be associated with NGS sufficiency. Although the oncologist’s reason for ordering NGS in individual patients was not collected in our data set, early-stage ordering is a reflection of many factors such as planned combined use of immunotherapy and chemotherapy, anticipation of future disease progression, and the need for treatment options in patients who are not surgical candidates. While NGS is standard in advanced NSCLC, recent trials have suggested variable benefit in early-stage disease as well [17,18,19].

We present the largest single center study of EBUS sampling with NGS. The yield we obtained with protocol adherence in our study (66%) aligns with yields reported in other studies (60–77%) including those by Hagemann et al. [11], Fielding et al. [20], and Rooper et al. [21]. In the study by Hagemann et al., of 381 patients with NSCLC, about 55% of samples were successfully sequenced [11]. When broken down by method of collections, endoscopic yields were found to be 66% (47 of 71 samples). Fielding et al. tested 66 EBUS samples of NSCLC and were able to successfully perform NGS testing in 60.1% (n = 40) [20]. Rooper et al. had a 76.6% success rate on 107 EBUS samples [21]. A systematic review and meta-analysis by Zhao et al. in 2022 reported a higher average yield, but the selection criteria for samples in some of the included studies makes it difficult to compare directly with our study [12]. The meta-analysis examined 21 studies comprising 1175 patients and found a pooled proportion of sufficient EBUS samples for NGS of 86.5% but some of these studies pre-selected samples for adequacy before attempting NGS testing [12]. For example, Stoy et al. reports a 95% success rate but samples were selected from a pathology database [22]. Likewise, Turner et al. assessed 315 NSCLC of which 200 were not tested and 115 were attempted for NGS testing [23]. They had an 86% success rate, but NGS testing was performed in-house where initial processing may preserve cells for NGS. Interestingly, 17 of those were pre-determined to be “borderline” for possible adequacy, suggesting that some samples excluded from NGS testing for perceived inadequacy might be able to yield results. Overall, pre-selection, exclusion of “borderline” results, and the use of in-house NGS processing may all contribute to higher NGS yields in these latter mentioned studies compared to ours. The substantial increase in yield we observed when our protocol was adhered to compared to non-adherence, supports its application and widespread use, although further studies are needed to optimize the protocol in hopes of improving yields even further.

The limitations of this study include its retrospective nature and its reliance on potentially incomplete documentation. NGS sequencing was not attempted on 84 samples and may be underpowered to detect size differences in yields. Not testing may be due to Oncology not requesting sampling (either because it was not needed or patient did not consent to testing or to treatment). Furthermore, reasons for deviation from the sampling protocol was not collected but may have included items such as difficult aspiration, bleeding, or poor patient tolerance of the procedure (thus limiting the time allowed for collection), or other clinical decompensation event (such as hypoxia). However, no difference was found in protocol adherence for moderate sedation versus general anesthesia (57.4% vs 35.7%, p = 0.27). If patient factors are the limiting agent for protocol adherence, then improved protocols may not result in improved study outcomes or tissue sample yields; however, there were few complications so patient factors are likely not an impactful limitation.

The strengths of this study include its robust sample size spanning a 6-year period. It fills a gap in the literature in regard to NSG yields and prediction factors, an area where prior studies are scarce and with smaller samples sizes.

When choosing biopsy locations for molecular genetics and NGS, proceduralists should consider aiming for larger lesions and obtaining three or more aspirations for NGS. Future studies comparing NSG yields between various sampling protocols, including reflex testing, should be considered to determine the ideal method for collecting and processing tissue samples for genetic sequencing. Reflex testing for NGS with upfront processing of the cell block may reduce material waste from preparation and result in higher yields.