Diagnostic performance of narrow-band imaging bronchoscopy in assessing tracheobronchial cell morphological abnormalities

Abstract

BACKGROUND:

Narrow-band imaging (NBI) bronchoscopy increases the diagnostic accuracy of lung cancer by enhancing mucosal surface structure images. The accuracy of NBI bronchoscopy for morphological abnormalities of tracheobronchial cells is unknown. This study aims to determine the diagnostic performance of NBI bronchoscopy in assessing tracheobronchial cell morphological abnormalities in nontumor mucosal airways.

METHODS:

This diagnostic study was conducted at Persahabatan Hospital from March 2023 to January 2024 using consecutive sampling. Participants aged over 18 with central lesions underwent NBI bronchoscopy. NBI vascular patterns were evaluated as the index test, with the gold standard being tracheobronchial cell morphological abnormalities assessed through histopathological examination. Diagnostic metrics, including sensitivity, specificity, and area under the curve (AUC), were analyzed using IBM SPSS Statistics 26.0.

RESULTS:

Among the 99 analyzed subjects, 46 exhibited abnormal vascular patterns on NBI bronchoscopy. A tortuous pattern exhibited the highest sensitivity and specificity (42.8% and 91.5%, respectively), compared with dotted and abrupt-ending patterns. Specificity for dotted and abrupt-ending patterns was 90.1% and 85.9%, respectively. However, as a potential screening modality, each NBI vascular pattern demonstrated low sensitivity, ranging from 14.3% to 42.8%. Combining tortuous, abrupt-ending, and dotted patterns improved diagnostic performance, achieving an AUC of 71.0%, sensitivity of 75.0%, and specificity of 67.6%.

CONCLUSION:

NBI bronchoscopic vascular patterns are correlated with tracheobronchial cellular morphological abnormalities, demonstrating high specificity and moderate sensitivity.

Background

Lung cancer is the leading cause of cancer morbidity and mortality worldwide. Global studies indicate that it is the most common cancer, with an incidence of 2.5 million cases in 2022 and accounting for 12.4% of all cancers globally.[1] Lung cancer mortality remains the highest among all cancers, with 1.8 million deaths, representing 18.7% of all cancer-related deaths.[1] This high mortality rate is linked to the fact that lung cancer is typically diagnosed at an advanced stage. Surveillance, epidemiology, and end results data show that the survival rates for lung cancer are significantly lower in cases of distant metastasis than regional metastasis and non-metastatic cases, with survival rates at 3%, 16%, and 27%, respectively.[2]

Screening programs enable the diagnosis of lung cancer at an early stage, allowing for curative surgical treatment that produces higher survival rates than advanced stages.[3] Various early detection programs for lung cancer have been developed with the primary goal of reducing mortality rates.[4] Sputum cytology and chest X-rays were initially introduced as early detection methods, but they proved to be less effective.[5] Computed tomography (CT) scans have since been developed as a potential screening tool. Lung cancer screening studies indicate that chest CT scans can detect nodules larger than 4 mm and identify 30%–100% of early-stage lung cancers.[6,7,8] Furthermore, chest CT screening has also been shown to help reduce lung cancer mortality rates.[9,10] However, CT scans are unable to assess minimal or precancerous lesions in the airways.[11,12]

Tissue biopsy via bronchoscopy is an essential part of early lung cancer detection programs and the screening for cell morphology abnormalities.[13,14,15,16] For example, in squamous cell lung cancer, cell morphology can transform from hyperplasia, squamous metaplasia (SQM), and mild dysplasia to severe dysplasia, and carcinoma in situ before carcinoma develops.[17,18] Carcinoma in situ has a significantly higher risk of developing into invasive cancer compared to mild dysplasia.[19,20,21] The morphology of cellular damage is closely linked to prognosis, making early clinical detection of such damage crucial for improving lung cancer survival rates. Intraepithelial airway tumors are associated with a high 5-year survival rate of 70%.[22,23] In addition, microinvasive or preinvasive lesions are often still suitable for surgical treatment, offering favorable outcomes.[24] SQM is a bronchial epithelial cell abnormality in response to smoking or exposure to other irritants or carcinogens. While some cases of SQM can regress to normal, others may disrupt airway function or progress to malignancy.[25]

Flexible fiberoptic bronchoscopy is a recommended standard procedure for patients undergoing lung cancer diagnostics, with a high safety profile.[26] The sensitivity of bronchoscopy is relatively high, reaching 88% for large airway lesions and 36%–88% for peripheral lesions, depending on the instrument modality used.[27] Previous studies have reported that 29% of carcinoma in situ lesions and 69% of microinvasive tumors can only be detected using white light bronchoscopy (WLB) by experienced operators.[28,29] However, early cancer or precancerous lesions, such as dysplasia or carcinoma in situ, may not be detected by standard bronchoscopy imaging or WLB.[30,31]

Narrow-band imaging (NBI) is a bronchoscopy technique designed to detect pathological changes in the microvascular patterns of the mucosa and submucosa. It uses two narrow wavelength light beams, blue light with a wavelength of 390–445 nm and green light with a wavelength of 530–550 nm.[32] NBI bronchoscopy can enhance its resolution by providing a contrasting view of mucosal and submucosal vascularization. Previous studies have shown that NBI is superior in detecting precancerous and cancerous morphological lesions, although the quality of the studies is heterogeneous, and many are retrospective reviews.[33,34,35,36] This study aims to determine the diagnostic performance of NBI bronchoscopy in assessing tracheobronchial cell morphological abnormalities in non-tumor airway lesions. This manuscript is written following the standards for reporting diagnostic accuracy studies checklist.[37]

Materials and Methods

Study design and participants

This diagnostic study was conducted at Persahabatan Hospital from March 2023 to 2024. The consecutive sampling method was used.

Eligible participants were individuals over 18 years of age who had central lesions identified by thoracic CT scans and were scheduled to undergo bronchoscopy procedures for lung tumor diagnosis.[38] Subjects with poor performance status or those who deteriorated before the procedure were excluded. Individuals who were later found not to have cancer or whose anatomical pathology samples were nonrepresentative were also excluded. The estimated sample size required in this study was calculated based on the estimated sensitivity and the assumed proportion of tracheobronchial cell morphological abnormalities. The estimated sample size is 102 subjects.

n = sample size

Zα = standard normal deviate for type I error (α = 0.05), which equals 1.96

Sensitivity = the sensitivity of bronchoscopy using NBI in detecting tracheobronchial cell morphological abnormalities, assumed to be 80%

d = desired precision, set at 10%

P = the estimated proportion of tracheobronchial cell morphological abnormalities among lung cancer patients, assumed to be 60%.

Test methods and bronchoscopic biopsy procedures

The vascular patterns identified through NBI bronchoscopy were used as the index test, whereas the gold standard was determined by the presence of tracheobronchial cell morphological abnormalities, as assessed through histopathological examination. The terminology for NBI vascular patterns includes dotted, tortuous, or interrupted patterns or combinations of these. The histopathological criteria adhered to the WHO classification.[39]

Bronchoscopy preparation and biopsy procedures were carried out in accordance with the standard operating procedures at Persahabatan Hospital. Subjects were required to fast for at least 6 h before the bronchoscopy. Following a re-evaluation in the bronchoscopy suite, general anesthesia was administered, and intubation was performed using a laryngeal mask airway (LMA). Vital signs were monitored throughout the preparation, procedure, and recovery phases. A flexible bronchoscope, model BF-1TQ170, with a working channel diameter of 2.8 mm, was utilized. The bronchoscopy machine employed was the Olympus EVIS EXERA III, featuring a Xenon light source, model CLV-190.

The scope was introduced through the LMA, and WLB observations were conducted, including an examination of the vocal cords, trachea, right main bronchus and its branches, and left main bronchus and its branches. These standard bronchoscopy examinations were meticulously documented. Subsequently, NBI bronchoscopy was performed to evaluate the main airways, including the trachea, carina, right main bronchus, and left main bronchus. Observations during bronchoscopy were carried out by two pulmonologists: the primary operator and a second pulmonologist with equivalent competence and experience. The assessments were jointly reviewed and agreed upon by both pulmonologists; in cases of disagreement, the findings were further discussed with an expert. The vascular patterns observed were carefully recorded and documented.

Once the vascular pattern was identified, a forceps biopsy was performed on the targeted area based on the NBI bronchoscopy findings. In cases where multiple areas with NBI findings were present, biopsies were taken from the area demonstrating the greatest abnormalities or from the most central location. If no vascular patterns were observed, biopsies were obtained from the normal carina or trachea. A minimum of five forceps biopsies were performed. Biopsies were not conducted on nodular infiltrative tumor masses suggestive of malignancy, lesions at least 1.5 cm from the infiltrative mass, or edematous mucosa. Single-use Boston Scientific Radial Jaw™ 4 forceps were utilized for the airway biopsies. The biopsies were carried out in real time with continuous monitoring via a display screen. The biopsy specimens were placed in 10% formalin for subsequent hematoxylin-eosin (H and E) staining. The diagnosis of tracheobronchial cell morphological abnormalities was determined by a pathology specialist based on an H and E examination. There was a blinding between the bronchoscopist and the pathologist.

Data analysis

Demographic and clinical data were obtained through direct anamnesis. Subjects were classified as heavy smokers if they had a Brinkman index >600. The Brinkman index was calculated by multiplying the average number of cigarettes smoked per day by the number of years the individual had been smoking. The study data were recorded and analyzed using IBM, NY, USA, SPSS Statistics 26.0. Diagnostic analysis was evaluated using a 2 × 2 diagnostic table approach. The diagnostic metrics assessed included sensitivity, specificity, positive predictive value, and negative predictive value, along with comparative analyses of the area under the curve for NBI vascular patterns.

Results

Clinical and bronchoscopy characteristics of the participants

We included 102 subjects and analyzed 99 of them. Three subjects were excluded from the analysis due to nonrepresentative histological samples, as shown in the Flow of Participants [Figure 1]. The average age of the subjects was 60 years (range: 38–81 years). Most subjects were male, heavy smokers, and had metastasis. The group with abnormal cell morphology in normal airways was associated with heavy smoking (odds ratio [OR]: 2.4; 95% confidence interval [CI]: 1.0–5.9; P = 0.02) and squamous cell carcinoma (SCC) pathology (OR: 2.9; 95% CI: 1.2–7.5; P < 0.01) [Table 1].

Figure 1.

STARD flow diagram of the study.

Table 1.

Clinical characteristics of participants in the study (n=99).

Characteristics	Abnormal cell morphology, n (%)	Normal, n (%)	OR (95% CI)	P
Male	22 (78.6)	46 (64.8)	1.9 (0.7–5.5)	0.18
>65 years old	10 (35.7)	18 (25.4)	1.6 (0.6–4.2)	0.30
History of tuberculosis	3 (10.7)	15 (21.1)	0.4 (0.1–1.7)	0.23
History of COPD	1 (3.6)	3 (2.8)	1.3 (0.1–14.7)	1.00
Heavy smokers	17 (60.7)	25 (35.2)	2.4 (1.0–5.9)	0.02
History of cancer in the family	6 (21.4)	8 (11.3)	2.1 (0.7–6.9)	0.21
SCC type	13 (46.4)	16 (22.5)	2.9 (1.2–7.5)	0.02
Metastasis	18 (64.3)	48 (67.6)	0.9 (0.3–2.2)	0.75
Superior vena cava syndrome	7 (25.0)	19 (26.8)	0.9 (0.3–2.5)	0.86
NBI vascular pattern
Abrupt-ending	4 (14.3)	10 (14.1)	1.0 (0.4–2.5)	1.00
Tortuous	12 (42.9)	6 (8.5)	3.4 (1.9–5.8)	<0.01
Dotted	5 (17.9)	7 (9.9)	0.8 (0.5–1.3)	0.31

COPD=Chronic obstructive pulmonary disease, NBI=Narrow band imaging, CI=Confidence interval, SCC=Squamous cell carcinoma, OR=Odds ratio

Of the subjects who underwent NBI bronchoscopy, 46 were found to have abnormal vascular patterns, while 2 were excluded due to non-representative samples. Abnormal NBI vascular patterns were most frequently visualized in the left main bronchi (46.5%), followed by the right main bronchi and trachea (36.4% and 17.2%, respectively). Twenty-one out of 44 subjects with vascular pattern abnormalities also showed abnormal cell morphology. The vascular pattern abnormalities observed included abrupt-ending, tortuous, and dotted patterns, as shown in Figure 2. There was no disagreement between the two evaluators regarding the vascular pattern assessment in all cases.

Figure 2.

Vascular patterns of narrow-band imaging bronchoscopy, (a) Normal vascular pattern, (b) Abrupt-ending pattern, (c) Tortuous pattern, (d) Dotted pattern.

This study demonstrates a high level of interobserver agreement in the evaluation of NBI bronchoscopic vascular patterns. Overall, the kappa value for agreement between observer one (a pulmonary specialist) and the primary investigator was 0.87 (standard error 0.06; P < 0.01), whereas the kappa value for observer two (a consultant pulmonary specialist in interventional pulmonology) was 0.90 (standard error 0.05; P < 0.01). The interobserver agreement for various vascular patterns, including tortuous, interrupted, dotted, and normal, was strong, with kappa values exceeding 0.80. These findings indicate a very good to near-perfect level of concordance between observers in identifying bronchoscopic vascular patterns using NBI.

Diagnostic performance of the narrow-band imaging vascular pattern

Table 2 presents the diagnostic performance of NBI bronchoscopy vascular patterns in detecting abnormalities in tracheobronchial cell morphology in non-tumor lesions. Various vascular patterns demonstrated different sensitivities with relatively high specificity. The tortuous pattern exhibited the highest specificity at 91.5%, compared with other vascular patterns. Dotted and abrupt-ending patterns also showed good specificity, with values of 90.1% and 85.9%, respectively. However, as a potential screening modality, each NBI vascular pattern demonstrated low sensitivity, ranging from 14.3% to 42.8% [Table 3].

Table 2.

Multivariate analysis of clinical factors and bronchoscopic vascular patterns associated with morphological abnormalities of tracheobronchial cells in lung cancer subjects (n=99).

Clinical characteristics and NBI vascular pattern	OR (95% CI)	P
Tortuous NBI vascular pattern	21.89 (6.50–73.71)	<0.01
Heavy smokers	5.39 (1.62–17.71)	0.01

NBI=Narrow band imaging, CI=Confidence interval, OR=Odds ratio

Table 3.

Diagnostic performance of narrow band imaging vascular patterns.

NBI vascular patterns	TP (n)	FP (n)	FN (n)	TN (n)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Abrupt-ending	4	10	24	61	14.3	85.9	28.6	71.7
Tortuous	12	6	16	65	42.8	91.5	66.7	80.2
Dotted	5	7	23	64	17.8	90.1	41.7	73.6
Any of tortuous, abrupt-ending and dotted	21	23	7	48	75.0	67.6	47.7	87.3

NBI=Narrow band imaging, TP=True positive, FP=False positive, TN=True negative, FN=False negative, PPV=Positive predictive value, NPV=Negative predictive value

The combination of tortuous, abrupt-ending, and dotted patterns produced the best diagnostic area (receiver operating characteristic [ROC] %) at 71.0% (95% CI: 60.0–83%) compared with the individual vascular pattern criteria, as shown in Figure 3. The sensitivity for the combination of these patterns was 75.0%, with a specificity of 67.6%. The positive predictive value for this criterion was 47.7%, indicating that nearly 1 in 2 subjects who met these criteria displayed changes in tracheobronchial cell morphology.

Figure 3.

Receiver operating characteristic curves of different narrow-band imaging vascular patterns for diagnosing abnormal cell morphology. Narrow-band imaging vascular pattern. Reference line, Tortuous pattern (0.67; 0.54–0.80), Dotted pattern (0.54; 0.41–0.67), Abrupt-ending pattern (0.50; 0.37–0.63), Any tortuous, abrupt-ending, and dotted pattern (0.71; 0.60-0.83).

Discussion

Until now, studies on vascular patterns using NBI have remained limited, with most research focusing on evaluating vascular patterns in mucosal lesions or endoluminal tumors. In this study, standard bronchoscopy was initially performed, followed by NBI bronchoscopy. However, this study focused on evaluating and performing biopsies on normal-appearing mucosa.[34,40] The vascular terminology used in this study follows that of previous NBI bronchoscopy studies.[34,41,42] The vascular patterns observed included interrupted, tortuous, and dotted vascular patterns. We did not encounter any nonspecific vascular patterns that could not be classified. Other studies have reported nonspecific patterns in addition to the three vascular patterns identified in this study.[42] Gastrointestinal NBI endoscopy clearly displays vascular patterns and allows for various classifications associated with precancerous and cancerous lesions, but this was not the case in this study.[43]

Most of the morphological abnormalities identified in this study were mild, with some exhibiting severe morphological changes. Severe dysplasia and carcinoma in situ were not found upon histological examination. In general, morphological abnormalities can be assessed using routine H and E staining in anatomical pathology without requiring additional immunohistochemical (IHC) tests. Cavarga et al. used IHC markers CD34, COX-2, and Ki67 on precancerous lesions to explore proliferative processes but not to determine the types of cellular morphological abnormalities.[44] Ideally, evaluating cytological changes, cell maturation, and architectural alterations in dysplasia requires a complete tissue sample on the slide, including the cilia, epithelial layers, basement membrane, and stroma. Incomplete tissue components can complicate and bias the diagnosis of cellular morphological abnormalities.

The relationship between cellular morphological abnormalities and NBI bronchoscopic vascular patterns remains under investigation. Previous studies have reported a correlation between vascular patterns and cellular morphological abnormalities, particularly in cases, in which vascular loops are present within the epithelial structure known as angiogenic squamous dysplasia (ASD).[45] In addition to ASD, Zaric et al. found a correlation between metaplasia and dysplasia and NBI vascular patterns, although the diagnostic criteria were not specified.[46] This study identified a positive association between tortuous vascular patterns and tracheobronchial cell morphological abnormalities. Shibuya et al. reported that abnormal tortuous NBI vascular patterns were associated with carcinoma in situ and malignancy.[45] Larger vascular diameters were correlated with a higher risk of malignancy. This finding aligns with our study, which observed that tortuous vascular patterns on visual bronchoscopy had larger diameters than normal vascular patterns.

Detecting cellular morphological abnormalities to assess cancer risk and screening for lung cancer presents a challenge for lung cancer control and management programs. Based on the ROC, we found that the diagnostic criteria of any of the tortuous, abrupt-ending, and dotted vascular patterns were the most effective in detecting cellular morphological abnormalities. Although the combination of tortuous, abrupt-ending, and dotted patterns showed higher sensitivity compared to other vascular patterns, the positive predictive value and specificity remained low. Factors such as malignancy, inflammation, and infection can influence the presence of NBI vascular patterns. In a meta-analysis, Elhidsi et al. reported sensitivity and specificity of 98% (95% CI: 89%–100%) and 77% (95% CI: 62%–88%), respectively, for NBI bronchoscopy in diagnosing malignancy in tumor lesions.[45] In another study, Herth et al. reported that NBI bronchoscopy detected 36 cases of precancerous metaplasia out of 40 bronchoscopy procedures, although biopsies were performed on suspected malignant lesions, whereas our study focused on nontumorous lesions.[34] Vincent et al. conducted biopsies similar to ours and found that 13 out of 26 biopsies were precancerous lesions.[33]

The role of bronchoscopy in the early detection of lung cancer has progressed, transitioning from navigational bronchoscopy to robotic bronchoscopy, significantly enhancing the accuracy of cancer detection.[47,48] However, the challenge persists in accurately detecting cellular morphological abnormalities, particularly metaplasia and dysplasia, rather than the cancer itself. By identifying these morphological abnormalities, which cohort studies have indicated as risk factors for cancer development, bronchoscopy could potentially be utilized as a clinical screening tool for lung cancer, especially in cases involving central airway abnormalities.

Bronchoscopic modalities that evaluate tracheobronchial cellular morphological abnormalities are highly valuable for understanding the morphogenesis of various central lesion cancers. To date, the morphogenesis of SCC has been extensively studied, and its progression from normal cells to carcinoma has been demonstrated. However, the morphogenesis of central-type adenocarcinoma remains unclear. Recent studies have indicated that adenocarcinoma is the most prevalent type of central lung cancer.[38] Repeated bronchoscopic evaluations over time enable clinicians to assess individual airways and perform repeat biopsies. Morphological changes provide additional insights into the morphogenesis of tracheobronchial cells.

This study demonstrates the performance of NBI bronchoscopy in detecting morphological cell abnormalities, some of which may represent precancerous lesions. The ideal screening target population includes individuals at risk of lung cancer who exhibit no respiratory symptoms or only mild symptoms and show no radiological evidence of malignancy.[49] Based on our research, we identified a risk factor that should be considered for targeted screening: individuals with a family history of malignancy. The American Cancer Society’s lung cancer screening guidelines recommend screening individuals aged 50–80 who are current smokers or who have a 20-pack-year smoking history and quit within the last 15 years.[50]

This study has several limitations. First, it did not further explore the relationship between the vascular patterns observed through NBI bronchoscopy and angiogenesis, which could provide valuable insights into the biological mechanisms underlying tracheobronchial lesions and their potential malignancy. Moreover, this study did not conduct a detailed assessment of the histopathological characteristics of tracheobronchial cell morphological abnormalities, potentially missing important findings that could enhance diagnostic accuracy. These limitations highlight the need for future research to enhance understanding of the diagnostic value of NBI bronchoscopy.

Conclusion

NBI bronchoscopic vascular patterns are correlated with tracheobronchial cellular morphological abnormalities. The diagnostic performance of these vascular patterns in detecting cellular morphological changes exhibits high specificity and moderate sensitivity. Further screening studies targeting high-risk populations are warranted.

Authors’ contributions

ME, JZ, LR, AA, AK, NNS, KT and MR contributed to the concepts, design, data collection and analysis; ME, JZ LR and AK contributed to statistical analysis and manuscript preparation. All authors reviewed and contributed to the final version of the article.

Ethical statement

The research proposal was approved by the Ethics Committee of Persahabatan Hospital, receiving ethical clearance under No. 13/KEPK-RSP/01/2023, and by the Ethics Committee of RSCM Faculty of Medicine, Universitas Indonesia, with ethical clearance No. KET-361/UN2.F1/ETIK/PPM.00.02/2023.

Declaration of Helsinki (For human research)

The principle of Helsinki Declaration for human research was considered in this study and the free informed consent has signed by all patients and all of them accept the medical procedures.

Patient consent

The written informed consent was obtained from all patients for this study.

Conflicts of interest

There are no conflicts of interest.

Data availability statement

Data are available if requested.

Funding Statement

This study was funded by the Directorate of Research and Development, Universitas Indonesia, Funder ID. No. NKB-664/UN2.RST/HKP. 05.00/2003, Awards/Grant Hibah PUTI 2023.