Narrow band imaging reveals field cancerisation undetected by conventional white light: Optical diagnosis versus histopathology

To assess whether narrow band imaging (NBI) detects fields of cancerisation around suspicious lesions in the upper aerodigestive tract, which were undetected by white light imaging (WLI).


| INTRODUCTION
Oncological work-up of Head and Neck Squamous Cell Carcinoma (HNSCC) patients comprises endoscopic procedures of the larynx and pharynx.The gold standard for diagnosis is histopathology, acquired through biopsy of the lesion during flexible laryngopharyngoscopy in the outpatient clinic or rigid endoscopy under general anaesthesia.
Endoscopic evaluation of mucosal lesions is usually performed by fiberoptic or high-definition videolaryngoscopy (HDL) using conventional white light imaging (WLI).The addition of narrow band imaging (NBI) shows more accurate detection of HNSCC, making this a promising technique. 1NBI uses a narrow bandwidth filter of blue (415 nm) and green light (540 nm) to identify superficial mucosal lesions based on vascular patterns. 2Distinct vascular patterns, known as intraepithelial papillary capillary loops (IPCL), have been linked to (pre) malignant lesions in different sites of the oral cavity, pharynx, and larynx. 3,4Nowadays, NBI has come to be a tool in the detection of primary and recurrent HNSCC.
Smoking and alcohol abuse potentially cause dysplasia and field cancerisation of the oral cavity, pharynx and larynx, contributing to the development of HNSCC. 5,6In cases of carcinoma in a wider field of premalignant changed mucosa, identification of tumour margins is important for tumour staging and treatment planning.Matsuba et al. found that, when using NBI compared to WLI only, more superficial high-grade dysplasia and invasive carcinoma was detected in the periphery of the primary tumour. 7Although the additional value in the detection of a primary untreated squamous cell carcinoma has already been established, the extent to which we may be missing cases of moderate or severe dysplasia or carcinoma (in situ) surrounding the carcinoma remains uncertain when using only WLI instead of combining it with NBI.

| Objective
We aimed to assess whether NBI detects fields of cancerisation around suspicious lesions in the upper aerodigestive tract, which were undetected by WLI. was taken.As a control, an additional biopsy of normal mucosa assessed by both WLI and NBI (WLIÀ/NBIÀ) was performed in a limited number of patients.
Demographic and clinical data of patients were collected.TNM staging was performed according to the 8th American Joint Committee on Cancer (AJCC) classification. 8Histopathological analysis was performed by a dedicated head and neck pathologist, blinded for the WLI and NBI optical results, and following WHO guidelines, describing lesions as either benign, mild, moderate or severe dysplasia, carcinoma in situ or invasive carcinoma. 9Severe dysplasia, carcinoma in situ, invasive carcinoma, and other malignant tumours were regarded as (pre) malignant, that is, as a positive result.All lesions described otherwise were considered as non-malignant, that is, as a negative result.Figure 1 shows representative WLI and NBI and histopathological images.

| Statistical analysis
SPSS version 23 statistical software (IBM Corp., Armonk, NY) was used to perform statistical analyses.All variables were tested for normality.Due to non-normal distribution of the data, the Mann-Whitney U test was used for comparing the means of two variables and the Kruskal-Wallis test was used in case of more variables.To evaluate the relation between optical diagnosis and histopathology, the Jonckheere-Terpstra test for trends was performed.If applicable, effect sizes were calculated to measure the strength of correlations, making it a critical component when valuing a statistical claim. 10fect sizes of 0.5-0.8 are considered to be moderate to large. 11

| RESULTS
Clinical data of 96 patients are presented in Table 1.Most lesions were located in the larynx (n = 47; 49%).In total, 206 biopsies were

Key points
• Narrow band imaging (NBI) is a promising technique for the accurate detection of head and neck squamous cell carcinoma (HNSCC) by identifying superficial mucosal lesions based on vascular patterns.
• NBI detected additional cases of moderate or severe dysplasia or carcinoma (in situ) surrounding the carcinoma that would have been missed by WLI alone.
• NBI-negative mucosa showed no signs of severe dysplasia, carcinoma in situ or malignancy, indicating its reliability in ruling out these conditions.

| Comparison of WLI+/NBI+, WLIÀ/NBI+ and WLIÀ/NBIÀ biopsies with histopathological diagnosis
As shown in Figure 2, 84 of 96 (88%) WLI+/NBI+ biopsies were (pre) malignant.Histopathological findings were significantly associated with the finding of a WLI+/NBI+ lesion ( p < .001).Of the 60 WLIÀ/ NBI+ biopsies, 19 (32%) were considered as (pre)malignant, that is, invasive carcinoma, carcinoma-in-situ or severe dysplasia, after histopathological examination.In another 16 cases (27%) of the 60 WLIÀ/ NBI+ biopsies, mild or moderate dysplasia was diagnosed (Table 3).WLIÀ/NBI+ directed biopsies detected carcinoma (n = 2), severe dysplasia, moderate dysplasia or mild dysplasia in 35 (58%) of the 60 tissue samples.None of the 46 biopsies optically assessed as normal mucosa (WLIÀ/NBIÀ) showed carcinoma or severe dysplasia, although three (6.5%)showed moderate dysplasia.In this study we compared the optical NBI guided diagnosis to the histopathological diagnosis.We demonstrated that NBI, compared to WLI only, is superior in detecting field cancerisation.The additional use of NBI revealed histopathologically verified (pre)malignant mucosal changes in 32% of the biopsies of with WLI normal appearing mucosa adjacent to the tumour.Moderate or mild dysplasia was found in an additional 27% of these biopsies.
In daily clinical practice, each head and neck oncologist has learned to interpret the optical information provided by the unfiltered spectrum of normal white light.With the introduction of filtering techniques such as NBI, the mind-set of the endoscopist must be changed for a correct optical diagnosis.Recent reviews on the clinical value of NBI advised a direct comparison of optical NBI diagnosis with the histopathological diagnosis. 12,13In this study, we assessed both WLI and NBI optical diagnoses with histopathological results of in procedure NBI-directed biopsies of the same lesions.Furthermore, we compared WLI+/NBI+ detected and WLIÀ/NBI+ detected suspicious tissue with negative controls.Our results indicate that NBI negative tissue shows no signs of severe dysplasia, in situ carcinoma or malignancy.
A study by Chabrillac et al. analysed WLI-/NBI+ lesions and found a 26% positive detection rate in at least one of the biopsies.In their study, carcinoma, moderate and severe dysplasia (regarded as high-grade dysplasia), and low-grade dysplasia were considered positive. 14In our study a detection rate of severe dysplasia or carcinoma was found in 32%, and when moderate dysplasia was considered as a positive result as well, an even higher detection rate of 58% was found.In the study of Chabrillac et al., 72 biopsies in 38 patients were performed, that is, more than one sample per patient, without explanation as to why more than one biopsy was performed and without specification of the exact location. 14The authors reported that mucosal samples were partly guided by NBI-pictures pre-operatively taken during flexible laryngoscopy in the outpatient clinic by other clinicians and biopsies were not always taken from the intended places.Furthermore, histopathology was not compared with NBI negative samples, and if lesions occurred to be non-malignant or if lesions could not be located properly, they were excluded.We eliminated these possible causes of bias by including patients with a lesion with a clinical suspicion of malignancy, performed biopsies with optimal WLI and NBI visualisation under general anaesthesia and took one WLIÀ/NBI + biopsy per patient.We compared histopathology of WLIÀ/NBI + biopsies with WLIÀ/NBIÀ samples as a negative control, as well as the WLI+/NBI+ samples.Additionally, we considered low-grade dysplasia as a negative result in WLIÀ/NBI+ detected mucosa, since the finding of low-grade dysplasia has no clinical consequences. 15,16A B L E 2 Ni-classification in relation to histopathology in 196 biopsies.There is no general upper aerodigestive tract classification of NBI detected abnormal blood vessels.The Ni-classification was developed for laryngeal lesions only, and NBI detection of abnormal blood vessels depends on the assessed mucosa type in the upper aerodigestive tract. 3,17To ensure a uniform determination of all lesions, we applied the Ni-classification to pharyngeal lesions as well.Six of the WLIÀ/ NBIÀ biopsies were determined as Ni type IV, but all were regarded as normal mucosa due to the location in tonsillar tissue or in lymphoid tissue at the base of tongue.These locations are known for showing false positive suspicious IPCLs on NBI visualisation. 18As discussed in other studies, NBI is a reliable tool for the detection of HNSCC, but sensitivity and specificity vary among several studies, mostly depending on the studied population. 19In our study, all 100 (pre)malignant lesions showed blood vessels of Ni-type IV or type V.Although we found a 100% sensitivity, the specificity of 42% was rather low when evaluating all biopsies.Both high sensitivity and low specificity are explained by the methodology of the study, in which we chose to take biopsies of WLI unsuspicious, but NBI suspicious mucosa at the border of the WLI suspicious mucosa, whereas most studies take biopsies of single lesions directed and detected by both WLI and NBI. 13 Although NBI has a high sensitivity for detecting malignant and benign lesions, the sensitivity and positive predictive value of detecting precancerous lesions are much lower. 20However, the definition of a precancerous lesion differs among studies, making a direct comparison problematic. 14,20In our study, only 22% of the lesions with Ni type IV blood vessels was diagnosed as either carcinoma or severe dysplasia, that is, (pre)malignant lesions.If we consider moderate and mild dysplasia as precancerous, then 44% of the Ni type IV lesions would be positive.This discrepancy in definitions underlines the need for careful interpretation of accuracy rates of both NBI and WLI.
Most notably, this study shows correlations between higher Niclassification and worse histopathology supported by moderate to large effect sizes between NBI-suspicious biopsies and histopathological outcome. 11No other studies tried to calculate these correlations and effect sizes; this would be helpful in comparing the results of various studies evaluating different populations.
When focussing on the additional value of NBI alone, we found that none of the NBI-/WLIÀ biopsies showed signs of malignancy.
This is an important finding since it proves that NBI-mucosa without Ni type IV or V blood vessels indeed can confirm the absence of severe dysplasia, carcinoma in situ or invasive carcinoma.After NBI examination of with WLI normal appearing mucosa, almost one third (32%) of the 62 samples contained (pre)malignant changes and even 58% when moderate dysplasia is considered as an early sign of intraepithelial neoplasia. 21This finding further strengthens the current position of NBI in determining the extension of the field of cancerisation and influence on staging of HNSCC. 19The superior value of NBI compared to WLI also raises the question whether NBI would further improve margin detection during transoral excision of HNSCC. 22However, the use of NBI alone, that is, without WLI, is not advised since NBI alone has a high chance of up to 70% in detecting non-malignant tissue, with a risk for unnecessary tissue removal, as already stated in earlier studies. 13Although using NBI did not lead to adjusted staging of the primary tumour in this study, these results indicate that upstaging based on additional NBI suspicion is conceivable.An upgrading of clinical T stage in 4.8%-12% of cases examined has been described.Future endeavours should focus on studying the value of NBI during endoscopy in post-treatment patients with a suspicion for residual or recurrent tumour under general anaesthesia.Interpretation of optical information of the irradiated mucosa is even more difficult compared to mucosa of untreated patients. 12In these cases, the correlation of histopathology with WLI and NBI optical diagnosis of tumour margins could be crucial for complete tumour resection in salvage surgery and directly influencing prognosis. 23 nant lesions, and the correlation between Niclassification and histopathology supported the accuracy of NBI diagnosis.performed.The Ni-classification could not exactly be determined in 10 lesions because of three reasons: either due to mucosal bleeding obscuring NBI visualisation (n = 3), the lesion being located in the tonsil or base of tongue (n = 6), or the lesion not showing any NBI blood vessels to apply the Ni-classification to (n = 1).
Histopathological data of all tissue samples in relation toNi type IPCL classification are shown in Table 2.All Ni type II lesions (n = 39) were non-malignant.Of the 41 Ni type IV lesions, 9 contained severe dysplasia or SCC (22%), 9 (22%) indicated moderate or mild dysplasia and 23 (56%) showed no form of dysplasia.In 91 out of 116 (78%) Ni type V lesions, histopathological analysis confirmed a (pre)malignant lesion: squamous cell carcinoma was diagnosed in 79 (68%), severe dysplasia in 10 (8.6%) and other malignancy in two (1.7%).Of Ni type V lesions, 25 of 116 (22%) were considered false positives, whereas this was the case in 32 of 41 (78%) Ni type IV lesions.All 15 biopsies with severe dysplasia showed Ni type IV-V blood vessels.

F
I G U R E 1 Representative examples of a 56-year-old male, presenting with a lesion in the left piriform sinus.(A) white light imaging (WLI) suspicious lesion located in the left piriform sinus.(B) narrow band imaging (NBI) image of that lesion, showing type V blood vessels.(C) Histopathological sample showing squamous cell carcinoma (haematoxylin-eosin staining, 400Â magnification).(D) WLI image of the same patient of an initially less suspicious lesion of the epiglottis.(E) NBI image of that lesion, arrows pointing towards type IV blood vessels.(F) Histopathological sample shows severe dysplasia (haematoxylin-eosin staining, 400Â magnification).

F I G U R E 2
Abbreviations: NBI, narrow band imaging; SCC, squamous cell carcinoma; WLI, white light imaging.
Demographic and clinical data of 96 patients.
a 1 lympho-epithial carcinoma and 1 neuro-endocrine tumour.T A B L E 1 Limitations of this study are the absence of a negative control WLIÀ/NBIÀ in 52% of the patients, but for ethical reasons we did not want to take biopsies of apparently normal mucosa in all patients (because of unfortunate localisation on the vocal fold).Another limitation is the extrapolation of the Ni-classification to non-laryngeal sites, nonetheless, we chose this approach to enable direct comparison of all biopsies.
| CONCLUSIONNBI is not only helpful in differentiating between normal and (pre) malignant mucosa, but also detects fields of (pre)malignant mucosal changes which would have been undiscovered if only WLI was used in 32% of the cases.
analysis; interpretation; manuscript editing.BECP: Study design; data acquisition; head and neck surgeon; data analysis; interpretation; manuscript editing.