Evaluation of narrow band imaging in the assessment of laryngeal granuloma

Laryngeal granulomas belong to common complications following trans-oral laser microsurgery (TLM). The aim of this study was to evaluate NBI in the differentiation between granuloma-like lesions and local tumor recurrence. 154 consecutive patients after TLM due to early laryngeal cancer were enrolled. In the group, a monthly follow-up including NBI endoscopy was performed. Moderate and severe dysplasia, carcinoma in situ and invasive cancer were defined as positive histology, laryngeal granuloma and other benign laryngeal lesions as negative histology and premalignant lesions as suspicious histology. In 47/154 (31%) cases, granuloma-like lesion (GLL) was found. Patients with GLL were divided into two groups based on the NBI classification. In all patients, the microvascular pattern in NBI was compared with the final histology. In group A, with suspicious, perpendicular vessels, 13/13 (100%) samples were positive. In group B, with normal vascular pattern 3/34 (9%) samples were positive and 31/34 (91%) samples were negative. There was a significant correlation between the positive NBI vascular pattern and the final histology (p = 0.00001). Sensitivity, specificity, accuracy of NBI were as follows: 81%, 100%, 94%, respectively.Based on our results, NBI can reliably differentiate between postoperative laryngeal granuloma and local tumor recurrence. In such a manner, this method is very helpful in the follow-up of tumor patients.


Material and Method
We enrolled 154 consecutive patients with confirmed diagnosis of laryngeal cancer, who underwent TLM cordectomy between April 2015 and November 2017 in two tertiary referral centres in Poland: Poznan 114/154 (74.03%) and Lodz (40/154 (26.97%) Medical University Department of Otolaryngology. Demographic and clinical characteristics of enrolled patients are presented in Table 1.
Following TLM, all patients were scoped once a month with trans-nasal flexible video-endoscope (Olympus Medical System Corporation, Tokyo, Japan) by means of white light, NBI and video-stroboscopic evaluation.
First, each patient with GLL underwent trans-nasal flexible video-endoscope (Olympus Medical System Corporation, Tokyo, Japan) with the optical filter for NBI. By using blue light (NBI), we assessed the vascular pattern of glottis and in particular vascular pattern of granulation tissue and mucosa surrounding the lesion. The diagnostic criteria defining suspected character of the lesion in NBI endoscopy were as follows: symmetrical arranged brown dot-like loops with narrow-angled turning point, which correspond to perpendicular vascular changes according to European Laryngological Society (ELS) classification 8 .
Detecting perpendicular vascular changes during NBI, in or surrounding granulation tissue, classified those patients to the "suspicious" group A. The remaining patients, presenting with longitudinal vascular pattern, which is considered normal, were qualified to the group B. Following NBI endoscopy, excisional biopsies with a cold instrument under general anaesthesia were performed in all patients. Eventually, the surgical specimens were sent to pathology. Moderate and severe dysplasia, carcinoma in situ and invasive cancer were defined as positive histology, laryngeal granuloma and other benign laryngeal lesions as negative histology and premalignant lesions as suspicious histology.
The main predictive variables were mucosa and submucosa vascular pattern in patients with GLL (according to ELS classification) and final histology. The additional variables were age, sex and number of days following TLM to GLL detection.
The primary outcome measure was the correlation between the observed vascular pattern of mucosa (according to ELS classification) surrounding GLL and the result of the final histology.

Result
Out of 47 patients, who presented with GLL, 13/47 (28%) (group A) had perpendicular vascular changes visualized in NBI on the mucosa surrounding granulation tissue (Fig. 1). The remaining 34/47 (72%) (group B) had longitudinal vascular changes on mucosa surrounding GLL (Fig. 2). In all patients, the microvascular pattern in NBI was compared with the final histology.
The local recurrence of laryngeal cancer after excisional biopsy was found in 1/47 (2%) patient during 10 th month of follow-up. In presented group, the mean number of days to detection of GLL was 164 days. The earliest detected GLL was visualized after 10 days and the latest after 577 days following TLM. There was no statistically significant correlation between presence of GLL and sex, smoking history, initial T-status, type of cordectomy or treating department (Table 1).

Discussion
Narrow band imaging has been in use in laryngology for 10 years now. In our previous papers, we proved that the NBI sensitivity, specificity and accuracy in detection of premalignant and malignant head and neck tumors is higher compared with classical endoscopy with white light only 9,10 . At first, NBI positive findings were identified as well-demarcated areas with scattered brownish spots 11,12 . In 2011, Ni et al. introduced the NBI classification for premalignant and malignant lesions in larynx. This classification consists of five types of different vascular patterns; additionally type 5 is divided into three subtypes. Types 1-3 are identified as vascular pattern characteristic for benign lesions, whereas types 4-5 are common in premalignant and malignant lesions 13 . However, the aforementioned classification is difficult to implement in everyday practice. Therefore, European Laryngeal Society (ELS) in 2016 proposed simpler classification for glottic lesions 8,14 . Authors distinguished only two types of vascular patterns. The first type is characterized by longitudinal vessels in two dimensions (length and width) and represents benign lesions. The second type, which are perpendicular vessels in third dimension towards the surface of the epithelium, represents premalignant and malignant lesions 8 .
In our study, the new classification introduced by ELS (2016) was used to assess vascular pattern surrounding granuloma-like lesions following TLM. Lesions characterized by perpendicular vascular pattern or intraepithelial papillary capillary loops, were identified as NBI positive, others as NBI negative.
Granulomas can be divided according to their etiology into two main subtypes: spontaneous granulomas, also known as contact granulomas, and iatrogenic granulomas 15 . In this study, we assessed only patients with iatrogenic granulomas, sequel of CO 2 laser cordectomies. The inflammation process that is associated with infection of the perichondrium is a trigger in granuloma formation process. According to many authors, the common histological features of granulomas are identified as nonspecific reparative granulation tissue covered by hyperplastic squamous epithelium with acanthotic thickening 16,17 . In this study, the definite confirmation of the benign character of granuloma-like lesion was the granulation tissue without atypical cells in the final histology examination.
Shoffel-Havakuk et al. 15 claimed that smoking is a significant risk factor in patients with iatrogenic granuloma. In our group, a significant proportion of patients 23/47(49%), who developed GLL were in fact heavy smokers. Schoffel-Havakuk et al. 15 considered a pedunculated character of lesion as a common feature of iatrogenic granulomas; we confirmed this finding in our study in majority of patients (40/47).  www.nature.com/scientificreports www.nature.com/scientificreports/ The main feature of all granulomas is their typical location: the posterior third of the glottis. The primary laryngeal cancer in this location is very rare. If present in this site, usually is a result of cancer extension from the anterior glottic lesion 18 . Among our patients, 22/47 had granuloma in the anterior commissure, 20/47 in the posterior third of the glottis and 5/47 in middle part of the glottis. NBI positive lesions (13/47), which were confirmed in histology, were localized respectively: 8 in anterior commissure, 3 in middle part of the glottis and 2 in the posterior part of the glottis. According to these results, localization not always is a good prognostic factor to predict benign character of granuloma following TLM. Therefore, NBI can be very helpful in determination of the potential local recurrence or benign granuloma.
Iatrogenic granulomas have shorter recovery time and lower recurrence rate in comparison to spontaneous granulomas. Therefore, based on our findings, we recommend for patients after TLM with typical features for contact granulomas and without pathological vessels in NBI, conservative treatment (antireflux therapy) and close observation. According to our experience, the need for biopsy among patients after TLM should be based on clinical assessment by white light and NBI endoscopy and initial response to antireflux medication.

conclusions
Based on our results, NBI can reliably differentiate between postoperative laryngeal granuloma and local tumor recurrence. We recommend that in cases of typical granuloma without pathological vessels in NBI, conservative antireflux therapy should be implemented. However, these patients should remain under close and regular tumor follow-up.