Introduction

Lung cancer screening programmes and the widespread availability of computed tomography (CT) of the chest have increased the detection of solitary pulmonary nodules [1,2,3]. Tools such as positron emission tomography/computed tomography (PET/CT) evaluation have decreased the need for unnecessary solitary pulmonary nodules resection [4]. Nevertheless, it has diagnostic limitations and might not be universally available [5]. The yield of conventional bronchoscopy in peripheral lung lesions (PLLs) is limited [6], as low as 14% for those in the outer periphery [7, 8]. Electromagnetic navigational bronchoscopy (ENB) provides a safe and higher-yield diagnostic procedure for the evaluation of PLLs [9,10,11,12]. Its yield can be significantly affected by the simultaneous use of other tools such as radial endobronchial ultrasound (r-EBUS), fluoroscopy, rapid on-site evaluation (ROSE), or whether the procedure is done under moderate sedation vs. general anesthesia [13, 14]. With over 30 studies reporting on endoscopic ENB yield, most of them have used one or more of the above supporting diagnostic tools and techniques [6, 10, 12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44]. Therefore, data on the stand-alone diagnostic performance of ENB are limited [39, 45, 46]. We sought to identify the performance of ENB as a stand-alone diagnostic technique and the effect of the various sampling techniques used during the procedure.

Methods

Observational analytical single-center study performed in a prospective institutional registry of patients with peripheral pulmonary lesions of the Fundación Jiménez Díaz University Hospital in Madrid, attached to the Autonomous University of Madrid with patients who were recruited to be included in the NAVIGATE study [10]. The primary goals of the study included the identification of variables predicting the yield of ENB as a stand-alone diagnostic technique. Secondary aims included ENB yield after an 18-month follow-up period, yield of individual tissue-sampling tools, and pneumothorax rate.

Eligibility criteria

The eligibility criteria included consecutive patients who underwent ENB for the work-up of PLLs and who had clinical and radiological data available during an 18-month follow-up period (spanning from July 2011 to October 2015). Exclusion criteria included difficulties tolerating moderate sedation, evidence of a visible endobronchial lesion, or a different ENB indication, such as fiducial marker placement.

Clinical variables

Demographics, smoking history, presence of chronic obstructive pulmonary disease (COPD) [47, 48], COPD severity [47], and previous lung metastasis from extrathoracic primary (extrathoracic cancer). Never-smokers and those who had quit for > 15 years were labelled as “non-smokers”. Nodule characteristics recorded included size in its three spatial axes [x, y, z], largest diameter on any axis [49], lobar location [16], fissure attachment (fissure adherence involving at least 1/3 of the lesion), pleural attachment, distance to the pleura, and bronchus sign [50]. A PET/CT standardized uptake value (SUV) of ≥ 3 was considered suspicious for malignancy.

ENB System

The system included the superDimension™ navigation system software version 6.0 (Medtronic, Minneapolis, MN), the Edge™ locatable guide, and the Edge™ 180° degree firm extended working channel. A therapeutic, flexible video bronchoscope with a 2.8-mm working channel was used in all procedures (Olympus; Tokyo, Japan).

Procedure and sample processing

ENB was performed in a standard fashion following the manufacturer’s protocol [51]. The lesions in which ENB was not diagnostic underwent chest CT-guided biopsy, endobronchial ultrasound (EBUS), or thoracic surgery.

ENB samples that yielded specific diagnoses (both benign and malignant), corroborated by surgical biopsy/resection results, clinical and imaging follow-up, or, in the case of benign disease, clinical and imaging assessments, were classified as diagnostic ENB. In contrast, ENB was categorized as non-diagnostic when it failed to provide a specific diagnosis. In cases of non-diagnostic ENB, we determined the final diagnosis through alternative procedures, including thoracic surgery, CT-guided biopsy, or linear EBUS (EBUS-TBNA). For patients who underwent surgery following a diagnostic ENB sample (i.e., therapeutic lung tumour resection), we based the final diagnosis on the histological examination of the surgical specimen. Additionally, we followed up patients for at least 18 months, during which clinical and imaging examinations were conducted to evaluate the consistency or discrepancy with the ENB results.

All patients had chest CT scan images obtained on Digital Images and Communications in Medicine format with a 512 × 512 resolution, a slice thickness of 1 mm, and an overlap of 0.8 mm. Images were uploaded using iLogic® software to create a three-dimensional road map. Sedation and monitoring during bronchoscopy were conducted according to the recommendations of the Spanish Society of Pulmonology and Thoracic Surgery [52] and the American College of Chest Physicians [53]. Topical anaesthesia was provided with lidocaine. Intravenous midazolam (median dose, 4.5 mg) and fentanyl (median dose, 100 µg) were used for moderate sedation. General anaesthesia, fluoroscopy, ROSE, or r-EBUS were not used during any ENB procedure.

Tools and techniques used during ENB included: suction catheter [24], transbronchial biopsy forceps (TBBx), cytology brush (CB), transbronchial needle aspiration (TBNA), bronchoalveolar lavage (BAL), and bronchial washing. Supplementary Table 1 provides individual technique details.

Sample size

We estimated that the study would require a minimum sample size of 160 patients to have a least 80 patients with the outcome (diagnostic ENB) in the more demanding scenario of 50% of patients with diagnostic ENB [54, 55]. These 80 patients would be sufficient to build a binary logistic regression multivariate model with 8 covariates (10 patients with the outcome per covariates).

Statistical analysis

Statistical analysis was performed using STATA version 17 software (STATA Corp., Texas, USA). Descriptive statistics for all continuous variables were summarised as means, standard deviations (SD), medians, and interquartile ranges. Frequency distributions and percentages were reported for discrete variables. The association between each variable and the diagnostic yield was analysed. For each outcome, associations with the corresponding set of variables were checked by χ2 or Fisher’s exact test (for categorical variables). Confidence intervals (CI), odds ratios (OR), and p-values were reported; two-tailed p values of less than 0.05 were considered to indicate statistical significance. A receiver operating characteristic (ROC) curve was plotted, and the area under the ROC curve was calculated. We evaluated the statistical significance of the trend in the use of combined techniques using the Cochran–Armitage statistical test.

Results

Demographic characteristics and key results

ENB was performed on 173 patients, but 13 patients were excluded according to eligibility criteria. The mean age was 67.9 years (SD:11), and 65% were male (104/160) (Table 1). While we did not collect specific procedure time data for individual patients, it is worth noting that, on average, each ENB procedure at our institution typically lasts approximately 90 min. The overall diagnostic yield was 69.4%, based on 18-month follow-up data. Sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV) were 59%, 100%, 45.6%, and 100%, respectively.

Table 1 Demographics, Lesions and Procedure Characteristics
Full size table

Univariate analysis of diagnostic yield

ENB yielded a diagnostic result in 111 out of 160 cases (69.4%). In the remaining 49 cases with non-diagnostic ENB, the diagnosis was established through thoracic surgery in 38 cases (77.5%), CT-guided biopsy in 5 cases (10.2%), Linear EBUS (EBUS-TBNA) in 3 cases (6.1%), and other procedures in 3 cases (6.1%) (Fig. 1).

Fig. 1
figure 1

Diagnostic algorithm based on 18-month follow-up data. Notes: ENB Electromagnetic navigational bronchoscopy, CT Computer tomography, EBUS Endobronchial ultrasound, EBUS-TBNA Linear EBUS

Full size image

The diagnostic yields in patients with and without a history of extrathoracic cancer were 32% and 68%, respectively (p = 0.001) (Table 2). In patients with perifissural lesions, the yield decreased 17% compared to patients without these lesions (56% vs. 73%; p = 0.054), and lesions located in the inferior lobes presented a diagnostic yield of 65%. 74.4% (119/160) received a diagnosis of malignant tumour, 23.1% of benign tumour (37/160) and 2.5% of infections (4/160). The diagnostic yield of malignant tumours of primary pulmonary origin was 59%, metastatic lesions of extrathoracic origin 22%, and benign tumours 100%.

Table 2 Bivariate analysis of Diagnostic yield ENB based on demographic variables, lesion characteristic and etiology
Full size table

Multivariate analysis of factors affecting diagnostic yields

We identified several factors increasing the diagnostic yield, including lesion size of ≥ 15 mm in the Z axis, presence of bronchus sign, smoking history, and age > 75. On the other hand, the yield was decreased by factors like history of extrathoracic malignancy, perifissural lesions and location in lower lobes (Table 3). A prediction equation was developed using these variables (Table 4). The area under the ROC curve of the predictive model for diagnostic success during ENB was 0.83 (95% CI: 0.75–0.90; p < 0.001) (Fig. 2). We found that when the PLLs were identified as metastatic, the diagnosis decreased the ENB yield by 85%, with an OR of 0.15 (95% CI: 0.03–0.54; p = 0.01). The malignancy rate was higher for SU ≥ 3 (Supplementary Table 2).

Table 3 Multivariate analysis factors affecting diagnostic yield
Full size table
Table 4 ENB yield prediction equation
Full size table
Fig. 2
figure 2

ROC curve of the predictive model for diagnostic success during ENB including the area under ROC curve

Full size image

ENB Diagnosis at 18-month follow-up

The diagnostic yield of ENB was 69.4% (111/160). Of the patients who obtained a false negative by means of ENB, 100% (49/49) were eventually diagnosed with a malignant tumour. By contrast, in the patients whose diagnosis was confirmed, they presented a non-malignant diagnosis of 36.9% (41/111) and a malignant diagnosis of 63.0% (70/111). The diagnostic algorithm after an 18-month follow-up is shown in Fig. 1.

Yield of bronchoscopic sampling techniques

CB and TBBx provided the highest yields, 51% and 53%, respectively (Table 5). Bronchial washing had the lowest yield (30%). The yield was around 66% when the suction catheter, TBBx, BAL, and CB were combined. Adding diagnostic techniques such as TBBx and TBNA increased the diagnostic performance and showed a statistically significant trend (p = 0.011; p = 0.045 and p = 0.002) (Fig. 3a and supplementary Fig. 1). Additionally, the combined use of TBNA or CB with other diagnostic techniques showed a significant increase in performance, with a significant trend (Fig. 3b).

Table 5 Diagnostic yield of techniques and tools used during ENB
Full size table
Fig. 3
figure 3

Diagnostic yield of the TBNA and CB with others sampling techniques. Notes: TBBx Transbronchial biopsies, BAL Bronchoalveolar lavage, CB Cytology brush, TBNA Transbronchial needle aspiration. *trend in the use of transbronchial needle aspiration with other combined techniques (Cochran–Armitage statistical test). ** trend in the use of cytology brush with other combined techniques (Cochran–Armitage statistical test)

Full size image

Safety and pneumothorax

Pneumothorax occurred in 7.5% (12/160) of the procedures, but only 2.5% (4/120) required drainage. In total, 51 TBNA were performed, of which 7.8% (4/51) had pneumothorax as a complication. There was no statistical association between pneumothorax and the sampling tool used (Supplementary Table 3). No individual factor increased the pneumothorax risk.

Discussion

We found a good diagnostic yield based on 18-month follow-up data and using ENB as a stand-alone diagnostic tool. This number is in the range of reported pooled yields (58.6%–84%) [6, 23, 29, 33, 35, 39, 40, 42, 45, 46, 56,57,58], it represents an average assessment of ENB. Our diagnostic yield is good when taking into account that our mean nodule diameter was smaller than those described in diagnostic ENB studies (Supplementary Table 4) [6, 23, 29, 33, 35, 39, 40, 42, 45, 46, 56,57,58].

Several studies are similar to ours, since they used only ENB for all of their cases [42, 45, 56,57,58]. For example, Bertoletti showed a yield of 77.4% but with a much larger nodule diameter (31.2 vs. 16 mm) [45]. Ozgul et al. examined ENB yield in 56 cases, but r-EBUS was used in 26 of those cases [39]. The yield for non-r-EBUS cases was 71.4%, which is similar to our study. Further, Raval and Amir reported an 83.3% yield with a relatively small lesion size (19.3 mm) [46]. Although only ENB was used, they utilised a tidal volume expiration mapped ENB (Veran Medical Technologies), which limits the comparison.

Some of our independent yield predictors have been previously identified in studies with multivariate analysis [14, 16, 18, 20, 22, 27, 49]. For example, smoking history increased our yield by sevenfold. Ost et al. also reported a smoking association, less strong [14]. The effect of smoking might be related to the localised distortion effect that it generates at the bronchial architecture level, triggered by the chronic inflammatory effect [47]. This could facilitate locatable guide access to nearby solitary pulmonary nodules. The presence of a bronchus sign on CT increased the yield by fourfold. Seijo et al. reported such an association with an eightfold increase in yield [42]. Our study confirms their findings, as a bronchus sign increased performance by 17% compared to patients without bronchial signs in our sample. Since bronchus sign was only present in about 50% of the patients, for cases without bronchus sign, it is our practice to make strategies, such as modifying the location of the locatable guidewire on at least two occasions, while obtaining transbronchial biopsies to avoid the “all-or-none” diagnostic phenomenon [14]. On this wise, samples are collected from various regions close to the lesion. We have yet to verify this fact, but we believe that it could mitigate the effect of not presenting the bronchus sign.

PLLs in the lower lobes decreased our yield by close to 80% [16, 22]. This could be explained by diaphragmatic movement during inspiration, with a difference in PLLs location of up to 2.5 cm. [22, 59] Lesion diameter of > 20 mm in its largest axis [14, 49] and > 30 mm18 correlated with higher diagnostic yield in other studies. For us, a lesion size of ≥ 15 mm in the Z axis increased the yield close to fivefold. We postulate that a larger lesion on the Z-axis might provide better endobronchial exposure. This requires additional validation. Perifissural lesions decreased the diagnostic yield; this might occur because of a reduction in bronchus size and a more tortuous airway pattern, limiting the advance of the locatable guide. Age of ≥ 75 increased our ENB yield by 4.8-fold. The rationale for this effect is unclear.

One of the clinical factors that negatively modify diagnostic yield is presenting a personal history of previous extrathoracic cancer before ENB [60]. Those patients had a higher percentage of metastatic PLLs compared to those without a history of extrathoracic cancer: 25% (13/51) vs. 4.5% (5/109), respectively. In turn, PLLs of metastatic origin decreased the yield by 85%, with an OR of 0.15. Two studies assessed a history of extrathoracic cancer previous to ENB and suspicion of metastatic PLLs as factors decreasing the diagnostic yield of ENB [22, 23]. We believe the reasons for the decrease in profitability are that most metastases of tumours at the lung level are due to hematogenous dissemination and, additionally, to the development of a metastatic niche that provides the adequate microenvironment for the implantation and growth of disseminated tumour cells [60, 61]. Tsuboi et al. [62] documented a significant difference in the bronchoscopy yield of peripheral lung lesions secondary to primary bronchogenic malignancies versus lung metastases, at 76.5% versus 29.1%, respectively. They found that bronchial airway exposure was present in only 5.1% of the metastases < 2 cm in size. Pulmonary metastases follow a hematogenous spread and are surrounded by non-malignant tissue (fibroblasts, neovasculature, inflammatory cells, and extracellular matrix) [60, 61]. Such dissemination patterns compared to those of bronchogenic carcinoma nodules, plus limited endobronchial exposure, might explain the lower yield observed [62, 63]. We believe that this fact is of great importance for ENB and at the level of bronchoscopy as a diagnostic technique in pulmonology. Finally, using the independent variables associated with the diagnostic yield, we generated a model to predict the diagnostic yield of ENB with good discriminating capacity (area under the ROC curve: 0.83). We plan to validate this model in a future prospective study.

We also assessed the diagnostic performance of the tools and techniques used during ENB [14]. Combination of multiple sampling techniques, particularly TBBx and TBNA, positively impacted the diagnostic yield and a diagnostic yield of 69% was reached when the most common individual techniques were used together: suction catheter + TBBx + BAL + CB + TBNA + bronchial washing (Table 5). Chao et al. also noticed a significant yield increase (18%) when TBNA was added to r-EBUS (78.4%) compared to TBBx and bronchial washing without TBNA (60.6%) [64]. In general, TBNA appears to be underutilised [14], even in cases with pleural distance of ≥ 10 mm. This is likely due to technical difficulties manoeuvring the needle in more distal locations and to concerns about a higher pneumothorax risk. However, in our study, its use was not associated with an increased incidence of pneumothorax. Most needle-associated pneumothorax risk have been extrapolated from CT-guided TTNA data (pneumothorax as high as 23%, up to ¼ requiring chest tube drainage) [65]. We believe that TBNA is a safe tool for lesions ≥ 10 mm from the pleura, as seen in our study and recently confirmed in the large multicentre NAVIGATE study [10].

Finally, since our study used stand-alone ENB under moderate sedation, it is possible that the routine use of additional diagnostic tools or general anaesthesia could increase the yield of ENB [14, 22, 29, 33]. For example, Eberhardt et al. reported an 88% yield for combined ENB + r-EBUS versus r-EBUS (69%) or ENB (59%) alone [22]. Our study bears several limitations, including its single-centre, retrospective observational nature, which exposes it to the risk of an unmeasured confounder and might limit the generalisability of the results. Further, our samples were analysed by the same pathologist. This can introduce bias to the diagnostic yield of various sampling techniques once an initial sample is diagnostic. We did not use fluoroscopy, r-EBUS, or ROSE, which could have potentially increased our diagnostic yield; therefore, our results apply mainly to studies not using such techniques.

Conclusion

Our findings show that ENB, as a stand-alone diagnostic technique using a multimodality sampling method under moderate sedation, has a good diagnostic yield, mainly in the presence of the bronchus sign and the use of TBNA, without increasing the risk of pneumothorax. Notably, we generated a predictive model for ENB diagnostic yield, which should be prospectively validated to provide more clarity regarding the optimal selection of patients undergoing ENB.