Background

Peripheral pulmonary lesions (PPLs) are lesions that are not visible by bronchoscopy, including no findings of endobronchial mass, extrinsic compression, sub-mucosal infiltration, or bronchial stenosis [1]. The clinical diagnosis of PPLs remains a challenge for physicians. Currently, PPLs can be diagnosed using various techniques, including bronchoscopy, computed tomography (CT)-guided transthoracic needle aspiration, and video-assisted thoracoscopic surgery. However, CT-guided percutaneous biopsies are associated with a high risk of complications such as pneumothorax and hemoptysis, whereas video-assisted thoracoscopic surgery is associated with greater trauma. Therefore, bronchoscopic examination is considered a safer diagnostic technique. Conventional diagnostic procedures for peripheral lung nodules and masses include transbronchial biopsy (TBB), bronchial brushing (BB), and bronchial washing (BW). Still, they tend to obtain low positive results, with approximately 18–62% of diagnostic yield [2].

To our knowledge, transbronchial needle aspiration (TBNA) was useful for sample mediastinal lymph node and stage lung cancer [3]. Previous research has demonstrated that conventional fluoroscopy-guided TBNA plays an important role in diagnosing PPLs [4]. With the recent introduction of advanced imaging-guided techniques such as endobronchial ultrasounds (EBUS), electromagnetic navigation, and virtual bronchoscopy, the diagnostic rate of PPLs has increased. EBUS-TBB helped detect PPLs, with a high diagnostic yield of 67–79% [5,6,7]. Diagnostic rates of EBUS-TBB tend to be lower when the probe is adjacent to lesions compared to within lesions [7].

TBNA may improve the diagnostic yield, especially when the EBUS probe is adjacent to lesions. A limited number of studies have obtained results in the diagnosis of EBUS-TBNA in PPLs ranging from 62.5 to 93.7%, and the diagnostic rates when the probe was adjacent to lesions ranged from 59.1 to 67.9% [1, 8,9,10,11,12,13,14,15,16,17]. However, the diagnostic role of EBUS-TBNA for PPLs has yet to be determined because many of the published studies show great heterogeneity in results. In addition, the lack of equipment and specialized skills in most hospitals have severely limited the clinical application of electromagnetic navigation and virtual bronchoscopy. Therefore, this meta-analysis aims to summarize the valuable articles and conduct a pooled analysis of diagnostic yield to illuminate the value of EBUS-TBNA.

Methods

Search strategy

Firstly, we searched the studies for available meta analysis that had figured out the diagnostic accuracy of EBUS-TBNA in PPLs. No meta analysis was found. To provide a reference for clinical work, we searched PubMed and Embase for relevant studies published from January 1, 2000 to December 30, 2021, clearing the diagnostic value of EBUS-TBNA in patients with PPLs using the following search terms: (“EBUS” OR “EBUS-TBNA” OR “TBNA” OR “endobronchial ultrasound” OR “endobronchial ultrasonography” OR “endobronchial ultrasound-guided” OR “endoscopic ultrasound” OR “transbronchial needle aspiration”) AND (“peripheral pulmonary lesions” OR “PPLs” OR “peripheral lung nodules” OR “peripheral lung masses”).

Study selection and data extraction

These articles were selected based on the PICO as follows. Population: individuals with peripheral pulmonary lesions; intervention: EBUS-TBNA used; comparison: other conventional bronchoscopy procedures used; and outcomes: the diagnostic rate of the EBUS-TBNA (primary outcome) and factors influencing diagnostic rate of EBUS-TBNA (secondary outcomes). We used the PICO to find a better method to clarify a diagnosis of PPLs. The criteria for inclusion were studies involving the diagnostic accuracy of EBUS-TBNA in PPLs. The exclusion criteria were as follows: (1) Studies including TBNA under the guidance of electromagnetic navigation and virtual bronchoscopy. (2) Studies evaluating the value of EBUS-TBNA in mediastinal lymphadenectasis and lung mass where the results were conflated for PPLs and mediastinal nodes. (3) Small sample sizes (less than 10 patients). (4) Abstracts, letters, review, editorials and case reports. (5) Articles were not in English.

Data extraction and statistical analysis

Two independent investigators (LY Lou and X Huang) firstly searched and screened the titles/abstracts to exclude irrelevant and duplicate studies. Then, we read a full text and selected studies based on the inclusion and exclusion criteria, with divergence resolved by consensus or another researcher (JW Tu). The following data were extracted: authors, year of publication, location, sample size, research design, average age of patient, lesion size, diagnostic yield and complications. Statistical analysis was carried out via R software (version 3.6.3) and R studio (version 1.3.1093). We calculated the rate using a Log transformation method. Heterogeneity was assessed by I2 test. When I2 statistics ≥ 50%, we used the random effects model, otherwise adopted the fixed effect model. The presence of publication bias was evaluated using the Egger test and the sensitivity analysis was planned by using subgroup analysis.

Quality assessment

The methodological quality of these articles in this meta-analysis were assessed by two reviewers using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAs-2) [18] tool.

QUADAS-2 is a tool to evaluate diagnostic study’s quality for risk of bias and concerns regarding applicability. The risk of bias was evaluated into four domains: patient selection, index test, reference standard, and flow and timing.

Results

A total of 425 citations were reviewed (Fig. 1), of which 7 studies involving 510 patients met our inclusion criteria [1, 8, 10,11,12, 16, 17]. The types of EBUS included were the convex linear probe-EBUS, and radial probe-EBUS, and most studies employed a 21G dedicated EBUS-TBNA needle. Some studies (3 out of 7) [8, 12, 16] enrolled patients with lung cancer suspicion, while the rest of participants had unknown masses. Malignant and benign diseases were diagnosed through EBUS-TBNA in 5 [1, 10,11,12, 16] and 2 out of 7 studies [1, 16], respectively. The diagnostic yield of EBUS-TBB was performed in 4 out of 7 studies [1, 8, 16, 17]. Combined EBUS-TBNA with EBUS-TBB and BB and/or BW was 5 out of 7 studies [1, 8, 10, 12, 16]. Comparing the diagnostic yield of EBUS-TBNA and EBUS-TBB when EBUS probe was adjacent to the lesion was done in 3 out of 7 studies [1, 8, 16]. Notably, there were no serious EBUS-TBNA-related complications.

Fig. 1
figure 1

Flow chart of the literature search and trial selection process

Full size image

The diagnostic yield of EBUS-TBNA in PPLs ranged from 62 to 93.8%, with a pooled accuracy of 0.75 (95% CI 0.67–0.83); a random-effects model was adopted due to I2 statistics equal to 79% (Fig. 2). Moreover, we conducted a subgroup analysis based on the forms of EBUS that demonstrated a higher TBNA yield when linear EBUS was used (0.83 (95% CI 0.70–1.00) versus 0.72 (95% CI 0.64–0.82) (Fig. 2). As previous study has found that tumor size is associated the diagnostic yields of EBUS-TBNA [1]. Due to the lack of specific masses size, we can only carry out a subgroup analysis based on the mean size of masses that revealed a relatively high rate when mean size less than 30 mm (0.81 (95% CI 0.60–1.00) versus 0.71 (95% CI 0.63–0.79). In prospective studies, the diagnostic rate was lower (181/262; 69%) than in retrospective studies (190/248; 80%), but this difference was not statistically significant (P = 0.14). EBUS-TBNA showed a higher diagnostic yield (0.75, 95% CI 0.67–0.83) than EBUS-TBB (0.63, 95% CI 0.48–0.77) (Fig. 3), whereas EBUS-TBNA combined with EBUS-guided bronchoscopy procedures (TBB and/or BW and/or BB) demonstrated the highest diagnostic yield (0.83, 95% CI 0.79–0.87) (Fig. 3). In cases of malignant lesions, EBUS-TBNA diagnostic rate was 0.79 (95% CI 0.72–0.88) (Fig. 4). The diagnostic rate of EBUS-TBB was higher when the EBUS probe was within lesions than when it was adjacent to lesions (Fig. 5). However, the yield of EBUS-TBNA did not differ significantly between the two groups (RR = 1.17, 95% CI 0.96–1.44) and the diagnostic rate of EBUS-TBNA was higher when the EBUS probe was adjacent to masses than EBUS-TBB (0.64, (95% CI 0.53–0.74) versus 0.46 (95% CI 0.19–072) (Fig. 6). Consequently, when the probe was adjacent to the masses, EBUS-TBNA was much more likely to produce a positive result than EBUS-TBB.

Fig. 2
figure 2

Diagnostic yield of EBUS-TBNA of the included studies in the subgroup of EBUS

Full size image
Fig. 3
figure 3

Diagnostic yield of EBUS-TBB compared with combined EBUS-TBNA with EBUS-TBB/BW/bronchial brushing. A) EBUS-TBB yield. B) combined EBUS-TBNA with EBUS-TBB/BW/bronchial brushing. 95% CI 95% confidence interval

Full size image
Fig. 4
figure 4

Diagnostic yield of EBUS-TBNA of the malignancy lesions

Full size image
Fig. 5
figure 5

Comparison of EBUS-TBB yield when EBUS probe was within to lesions (experimental events) and adjacent to the lesions (control events)

Full size image
Fig. 6
figure 6

Diagnostic yield of EBUS-TBNA compared with EBUS-TBB when EBUS probe was adjacent to the lesions. A EBUS-TBNA yield. B EBUS-TBB yield

Full size image

Heterogeneity analysis

Clinical heterogeneity may be reflected by the type of study (prospective or retrospective), the inclusion of patients of varying ages, the size of PPLs, the forms of EBUS technology, the size of the needles, and the use of additional guidance tools, as well as by the different endoscopists with varying levels of experience (Table 1). In addition, the statistical analysis revealed that the heterogeneity was also significant (I2 = 79%, p < 0.01). A subgroup analysis was conducted to investigate the origin of heterogeneity, which suggested that types of EBUS may explain a portion of the observed heterogeneity. Still, it remained unclear as heterogeneity persisted in studies with r-EBUS (I2 = 77%) (Fig. 2). Meanwhile, stratified by the size of PPLs may also explain a portion of the observed heterogeneity when the size of PPLs ≥ 30 mm, the heterogeneity was relatively low (I2 = 58%). But there was still great heterogeneity when the size of PPLs < 30 mm and unknown size (I2 = 87% and I2 = 90%).

Table 1 Main characteristics of selected studies on EBUS-TBNA
Full size table

Sensitivity and quality analysis

The Egger test indicated no publication bias in EBUS-TBNA yield (P = 0.07). The sensitivity analysis demonstrated that all estimated values fell within the confidence interval (Fig. 7); therefore, the analysis was considered stable and reliable. The inability of the sensitivity analysis to explain statistical heterogeneity suggests that other factors caused heterogeneity. The application of QUADAS-2 tool revealed an overall low methodological quality (Fig. 8).

Fig. 7
figure 7

The sensitivity analyses

Full size image
Fig. 8
figure 8

Quality analysis of the studies included in the meta-analysis

Full size image

Discussion

Our meta-analysis compiled the findings of published studies assessing the diagnostic value of EBUS-TBNA in PPLs. It demonstrated an acceptable overall diagnostic yield (0.75, 95% CI 0.67–0.83) by EBUS-TBNA in PPLs, suggesting that this technique could be utilized routinely to diagnose PPLs. EBUS-TBNA diagnostic rate in malignant lesions was 0.79. The highest rate (0.83) was observed with the combination of TBNA and conventional bronchoscopy (BB, BW, and TBB) guided by EBUS. EBUS-TBNA was more effective than EBUS-TBB in obtaining a positive diagnostic result when the EBUS probe was adjacent to the lesions (0.64 vs. 0.46). We believed that all available bronchoscopic procedures (BB, BW, TBB, and TBNA) should be used to achieve the best outcome for PPLs suspected of having malignant lung disease, especially in cases where the EBUS probe was adjacent to lesions.

Our study suggested that the EBUS-TBB yield was 0.63 (95% CI 0.48–0.77). A number of previous studies indicated the diagnostic rate of EBUS-TBB in PPLs. D.P. Steinfort. et al. [19] did a meta-analysis of EBUS-TBB in PLLs, whose diagnostic rate was 0.73. Masahide Oki et al. [20] showed the diagnostic utility of EBUS-TBB was 0.69. In our study, EBUS-TBNA had a higher diagnostic yield (0.75, 95% CI 0.67–0.83) than EBUS-TBB, mainly when EBUS probe was adjacent to the lesions (0.64 vs. 0.46). Similarly, previous observations also support our results. Kurimoto et al. [21] found that the yield of TBB when EBUS probe was adjacent to the lesions was very low (7%). However, the diagnosis rate using TBNA alone was not significantly different according to the location of EBUS probe [1]. The reason may be that TBNA needle can completely pierce and penetrate the bronchial wall and sample tissue [1]. Therefore, when EBUS probe was adjacent to the lesions, EBUS-TBNA had a higher possibility of obtaining positive results. However, further studies about EBUS probe location should be conducted to make up for the small sample size deficiency.

In addition, we found the diagnostic rate of malignant lesions was quite high. Additional subanalysis supports the previous findings that malignant lesions are associated with a better yield [4, 22,23,24]. Patient selection may explain this result, as many studies only enrolled individuals with lung cancer suspicion. Furthermore, for lung cancer, EBUS-TBNA could get sufficient samples for histological diagnosis and various molecular testing to determine the appropriate targeted therapy for the patients [12]. Unfortunately, TBNA did not have sufficient evidence to confirm the diagnosis of benign diseases (such as tuberculosis, sarcoidosis, or mycotic infection) immediately. Based on the facts, we proposed that EBUS-TBNA is a viable option for PPLs predisposed to malignant tumors. It should be used in conjunction with TBB for patients with inflammatory or infectious lesions.

Next, we sought to identify the clinical factors influencing the EBUS-TBNA diagnostic rate. Trisolini R. et al. [24] analyzed that a lesion size > 2 cm, malignancy, and location in the middle lobe were independent predictors of a positive TBNA result in PPLs. Mondoni M. et al. [4] concluded that CT bronchus sign, malignant disease, lesion size > 3 cm, and rapid on-site evaluation (ROSE) were major predictors of a higher TBNA yield. However, little focus was on identifying the risk factors associated with a high EBUS-TBNA yield. Chao TY. et al. [1] identified lesion size and characteristics as clinical predictors of a higher EBUS-TBNA rate. In contrast, some studies found no statistically significant differences in diagnostic yields based on tumor size [15, 16]. So we stratified the average tumor size by 30 mm and found that there was no statistical difference in the diagnostic rate of EBUS-TBNA. The diagnostic rate was higher when the size < 30 mm (0.81 vs 0.71). It is regrettable that the included articles did not list the diagnosis results according to the specific tumor size. Thus, it may be less rigorous. It's unconvincing and the causes of the result may be missing data about sizes in two articles [12, 17] and the mean nodule size. Therefore, additional prospective studies are crucial to determine the factors including patient’s characteristics (such as smoking history, gender, and age), lesion size, lesion characteristics, malignancy, location of the lesions, with or without ROSE, EBUS probe location, etc.

According to subgroup analysis, linear EBUS appeared to have a higher incidence than radial EBUS (0.83 vs. 0.72). The form of EBUS may predict a greater yield of TBNA. The better result may have been that linear EBUS was able to generate sector view images and permit real-time TBNA [25]. However, the samples of linear EBUS were relatively small, and additional research was required to determine the advantage of linear EBUS.

It should also be documented that our meta-analysis had several limitations. First, there existed significantly higher clinical and statistical heterogeneity. Diverse types of EBUS are capable of explaining a portion of the observed heterogeneity, but not all of it. High clinical and statistical heterogeneity may be attributable to study design, nodule size, operator skill, X-ray fluoroscopy, and guide sheath use. Second, there is selection bias in some studies. Some enrolled patients were suspected of having lung cancer, and one article [16] only showed the diagnostic rate of EBUS-TBNA in PPLs who was diagnosed with lung cancer. Third, insufficient articles were pooled.

Conclusion

Our results suggest that EBUS-TBNA is a useful and safe technique for diagnosing PPLs. In PPLs suspected of malignancy, EBUS-TBNA should be performed first, mainly when the EBUS probe is adjacent to the lesions. Undoubtedly, the combination of EBUS-TBNA and conventional bronchoscopy techniques would achieve a higher yield of PPLs.