European Archives of Oto-Rhino-Laryngology

https://doi.org/10.1007/s00405-021-06723-7

LARYNGOLOGY

Validation of the European Laryngological Society classification

of glottic vascular changes as seen by narrow band imaging
in the optical biopsy setting
Francesco Missale1,2   · Stefano Taboni3,4   · Andrea Luigi Camillo Carobbio1,5   · Francesco Mazzola6   ·
Giulia Berretti7 · Andrea Iandelli1,5 · Marco Fragale1,5 · Francesco Mora1,5 · Alberto Paderno7 · Francesca Del Bon7 ·
Giampiero Parrinello1 · Alberto Deganello7   · Cesare Piazza7   · Giorgio Peretti1,5

Received: 7 December 2020 / Accepted: 23 February 2021

© The Author(s) 2021

Abstract
Purpose  In 2016, the European Laryngological Society (ELS) proposed a classification for vascular changes occurring in
glottic lesions as visible by narrow band imaging (NBI), based on the dichotomic distinction between longitudinal vessels
(not suspicious) and perpendicular ones (suspicious). The aim of our study was to validate this classification assessing the
interobserver agreement and diagnostic test performance in detecting the final histopathology.
Methods  A retrospective study was carried out by reviewing clinical charts, preoperative videos, and final pathologic diag-
nosis of patients submitted to transoral microsurgery for laryngeal lesions in two Italian referral centers. In each institution,
two physicians, independently re-assessed each case applying the ELS classification.
Results  The cohort was composed of 707 patients. The pathologic report showed benign lesions in 208 (29.5%) cases, papil-
lomatosis in 34 (4.8%), squamous intraepithelial neoplasia (SIN) up to carcinoma in situ in 200 (28.2%), and squamous cell
carcinoma (SCC) in 265 (37.5%). The interobserver agreement was extremely high in both institutions (k = 0.954, p < 0.001
and k = 0.880, p < 0.001). Considering the diagnostic performance for identification of at least SIN or SCC, the sensitivity
was 0.804 and 0.902, the specificity 0.793 and 0.581, the positive predictive value 0.882 and 0.564, and the negative predic-
tive value 0.678 and 0.908, respectively.
Conclusion  The ELS classification for NBI vascular changes of glottic lesions is a highly reliable tool whose systematic use
allows a better diagnostic evaluation of suspicious laryngeal lesions, reliably distinguishing benign ones from those with a
diagnosis of papillomatosis, SIN or SCC, thus paving the way towards confirmation of the optical biopsy concept.

Keywords  Narrow band imaging · European Laryngological Society classification · Optical biopsy · Laryngeal cancer ·
Vascular changes · Endoscopy

Francesco Missale and Stefano Taboni have equally contributed to

the manuscript.

Cesare Piazza and Giorgio Peretti have equally contributed as

senior authors.

4
* Andrea Luigi Camillo Carobbio Department of Clinical and Experimental Sciences,
[email protected] University of Brescia, Brescia, Italy
5
1 Department of Surgical Sciences and Integrated Diagnostics
IRCCS Ospedale Policlinico San Martino, Genoa, Italy
(DISC), University of Genoa, Genoa, Italy
2
Department of Molecular and Translational Medicine, 6
Department of Otolaryngology, Head and Neck Surgery,
University of Brescia, Brescia, Italy
IRCCS Regina Elena National Cancer Institute, Rome, Italy
3
Section of Otorhinolaryngology, Head and Neck Surgery, 7
Unit of Otorhinolaryngology, Head and Neck Surgery, ASST
Azienda Ospedaliera di Padova, University of Padua, Padua,
Spedali Civili di Brescia, University of Brescia, Brescia, Italy
Italy

13
Vol.:(0123456789)
 European Archives of Oto-Rhino-Laryngology

Introduction of vocal fold vascular changes as seen under NBI [12]. In

this system, the authors distinguished between normal and
Early detection and diagnosis of laryngeal squamous cell pathologic vascular patterns of the vocal folds. The latter,
carcinoma (SCC) are crucially involved not only in reducing in turn, were divided into longitudinal and perpendicular
mortality, but also to optimize therapeutic approaches aimed vascular changes. Longitudinal vessels characterize benign
at achieving the best organ and functional preservation [1, lesions, while perpendicular ones (i.e. dot-like intrapapillary
2]. Fortunately, glottic SCC, the most common laryngeal capillary loops [IPCL], enlarged and worm-like vessels) are
tumor localization, is more frequently detected at an earlier considered signs of benign neoplasms (such as papilloma-
stage than tumors originating in other subsites of the head tosis), squamous intraepithelial neoplasia (SIN), or frankly
and neck due to its early (albeit highly non-specific) symp- malignant lesions.
toms [3]. Laryngeal SCC examination is usually performed The present study aims to assess the performance of the
by flexible (video)endoscopy under white light (WL) and ELS classification of vascular changes in a broad multicenter
relies on the analysis of superficial characteristics (size, cohort, testing its interobserver agreement as primary end-
color, location, single or multifocal appearance) and visible point, and analyzing its accuracy in predicting the final path-
morphological features (smoothness, irregularity, keratini- ological results in an optical biopsy setting, i.e. by evaluat-
zation, ulceration, submucosal growth), per se non-pathog- ing the diagnostic accuracy of NBI by comparing it with
nomonic and possibly overlapping each other in malignant the final histopathologic diagnosis obtained after complete
and benign pathologies, especially when diagnosed at early removal of the glottic lesion.
stages. This implies the frequent need to obtain an incisional
biopsy before deciding on the therapeutic approach, with an
increase in costs, anesthetic risks, and potential undue dam- Methods
age to the vocal cords.
Computed tomography and magnetic resonance imaging Study population
definitively play a major role in diagnosis of more advanced
diseases, providing information about the involvement of A retrospective study was carried out enrolling patients
laryngeal structures and lateral neck lymph nodes, while, on treated at the Departments of Otorhinolaryngology—Head
the other hand, they fall shortly in identifying and character- and Neck Surgery of the Universities of Genoa (Center
izing superficial mucosal lesions. By contrast, the judicious A;  from January 2012 to December 2016)  and Brescia
use of high definition and better contrasted videoendoscopic (Center B; from January 2015 to December 2018), Italy.
images now offer staggering details in evaluation of epithe- All patients enrolled were affected by laryngeal lesions; a
lial and superficial vascular patterns. Moreover, clinicians pre-treatment videoendoscopic evaluation with both WL and
can increasingly benefit from novel optical diagnostic meth- NBI was performed in the office as well as in the operatory
ods, providing information even closer to those obtained by theater, and the records were saved in his/her medical chart;
formal histopathological examination, thus differentiating the laryngeal lesion was treated by a transoral microsurgical
between normal mucosa and discrete lesions and, among approach by either cold instrumentation and carbon dioxide
the latter, between those with benign versus malignant ­(CO2) laser; postoperative histopathologic assessment was
behaviors [4]. In this context, narrow band imaging (NBI) obtained to classify the resected tissue as benign, dysplastic
is a well-established bioendoscopic technique using filtered or malignant. Histopathological diagnosis was performed
wavelengths to enhance microvascular alterations associ- according to the WHO classification system [13].
ated with preneoplastic and neoplastic transformation of
the upper aerodigestive tract (UADT) mucosa [5–8]. Since Clinical diagnostic work‑up
its first introduction in the late 1990s, the use of NBI has
considerably upgraded physicians’ ability for non-invasive All patients were preoperatively evaluated by high-defi-
detection and delineation of suspicious mucosal lesions, nition television (HDTV)-WL and HDTV-NBI through a
and is thus beneficial in the diagnosis of a variety of benign videorhinolaryngoscope ENF-VQ or ENF-VH coupled to
and malignant lesions [9]. However, the need for a common an Evis Exera II HDTV camera connected to an Evis Exera
language to be shared among clinicians to describe NBI- II CLV-180B light source (Olympus Medical System Corp.,
enhanced vascular patterns led to the proposal of different Tokyo, Japan). Just before surgery, in the operating room,
classifications during the last decades [10–12]. In 2016, the with patient under general anesthesia, intraoperative HDTV-
Working Committee on Endoscopic Laryngeal Imaging WL and HDTV-NBI rigid endoscopy with 0° and 70° tel-
of the European Laryngological Society (ELS) published escopes (Karl Storz, Tuttlingen, Germany) was also system-
a new proposal for a simplified (dichotomic) description atically performed. On the basis of this diagnostic work-up,
laryngeal lesions were subsequently removed by either a

13
European Archives of Oto-Rhino-Laryngology

phonomicrosurgical approach (in case of benign lesions) or or other treatments before the index transoral microsurgi-
excisional biopsy (in case of papillomatosis, SIN, carcinoma cal procedure were retrieved from the hospital databases.
in situ [CIS] or invasive SCC) by type I–III cordectomies Two independent physicians from each institution with
according to the ELS classification of cordectomies [14]. at least a 3-year-experience in the use of NBI, blinded to
the final histopathologic result, retrospectively and inde-
Clinical evaluation applying the ELS classification pendently reviewed the intraoperative videoendoscopic
recordings. Applying the ELS classification for laryngeal
Clinical records of the study population, including demo- vascular changes [15], each case was categorized as sus-
graphic features and information on previous treatments picious for malignancy (presence of perpendicular vas-
in terms of laryngeal surgery, head and neck radiotherapy, cular abnormalities as shown in Fig. 1) or non-suspicious

Fig. 1  Endoscopic picture of three representative cases of SCC (a–d) or CIS (e, f) correctly identified as suspicious by the presence of perpen-
dicular vascular abnormalities (* in all panels) evaluating the NBI endoscopic appearance (b, d, f) and applying the ELS classification

13
 European Archives of Oto-Rhino-Laryngology

Fig. 2  Endoscopic picture of three representative cases of benign endoscopic appearance (b, d, f) and applying the ELS classifica-
glottic lesions: keratosis without atypia (a, b), Reinke’s edema (c, d) tion. The ° in all panels points to non-suspicious longitudinal vascu-
and polyp (e, f) correctly identified as benign lesions without iden- lar abnormalities that can be observed inside the lesion (d) of at its
tifying any perpendicular vascular abnormalities evaluating the NBI boundary (b, f)

13
European Archives of Oto-Rhino-Laryngology

Fig. 3  Endoscopic pictures of three representative cases (a, b; c, d; and e, f) of recurrent laryngeal papillomatosis correctly identified detecting
wide angle IPCLs (arrowheads in all panels) evaluating the NBI endoscopic appearance (b, d, f)

(undetectable perpendicular vascular changes or longitudinal papillomatosis (i.e., wide angle IPCL) was also considered
ones as shown in Fig. 2). In case of interobserver disagree- as a secondary endpoint (Fig. 3).
ment, consensus was reached by direct comparison between
the examiners. The identification of features of respiratory

13
 European Archives of Oto-Rhino-Laryngology

Table 1  Definition of interrater agreement qualitative scores accord- detecting papillomatosis-like features, the Clinical Util-
ing to Altman et  al. [15, 16] and Clinical Utility indexes grading ity (CU) indexes were also derived, taking into account
according to Mitchell [18]
the measures of occurrence (sensitivity or specificity)
Agreement classification Clinical Utility Index classification together with the possibility of discrimination (positive
κ Strength of CU+Ve or CU− Grading [PPV] or negative predictive values [NPV]), and their
agreement Ve qualitative grading were judged accordingly [18, 19]
(Table 1). The Positive Clinical Utility Index (CU + Ve)
 < 0.21 Poor  < 0.49 Poor utility
is defined as sensitivity*PPV, and a high CU + Ve results
0.21–0.40 Fair 0.49–0.63 Satisfactory utility
should characterize “case finding” tests. By contrast,
0.41–0.60 Moderate 0.64–0.80 Good utility
good Negative Clinical Utility Index (CU−Ve), defined
0.61–0.80 Good 0.81–1.00 Excellent utility
as specificity*NPV, should be ideal for “screening” tests
0.81–1.00 Very good
[18, 19].
CU + Ve positive clinical utility index, CU−Ve negative clinical utility In all analyses, a two-tailed p value < 0.05 was consid-
index ered significant. GraphPad Prism (San Diego, CA, USA),
Stata (version 13.0, College Station, Texas, USA) and R
(version 3.6.2) were used for statistical analysis and ren-
Statistical analysis dering graphs.

Clinical data were reported as absolute and relative fre-

quencies. The reliability of the ELS classification was Results
assessed for each independent cohort measuring the
Cohen’s k statistic and the strength of agreement graded Clinical data
according to Altman et al. [16, 17], as reported in Table 1.
Considering the final consensus of the evaluation, we A total of 707 patients met enrolment criteria: 434 (61.3%)
assessed the performance of the diagnostic test for pre- had been evaluated and treated at the center A, and 273
diction of the final pathologic result (at least SIN1, up to (38.7%) at the center B. Five-hundred and fifty six (78.6%)
SCC) in an optical biopsy setting. For better understand- were males and 151 (21.4%) females, with a mean age of
ing of the clinical utility of applying the classification and 61.8 years (range 18–91). Four-hundred and seventy-eight

Table 2  Clinical features of the Variables All Longitudinal vessels Perpendicular

cohort vessels
n % n % n %

All 707 100.0 283 40.0 424 60.0

Gender
 Male 556 78.6 184 26.0 372 52.6
 Female 151 21.4 99 14.0 52 7.4
Previous treatments°
 No 478 67.6 221 31.3 257 36.4
 Previous surgery 174 24.6 45 6.4 129 18.2
 Previous RT 17 2.4 4 0.6 13 1.8
 Previous biopsy 54 7.6 8 1.1 46 6.5
Histology
 Benign 208 29.4 184 26.0 24 3.4
 Papillomatosis without atypia 34 4.8 0 0.0 34 4.8
 SIN1 46 6.5 24 3.4 22 3.1
 SIN2 61 8.6 26 3.7 35 5.0
 SIN3/CIS 93 13.2 15 2.1 78 11.0
 SCC 265 37.5 26 3.7 239 33.8

The sum of rows is 723 since 16 patients had two different previous treatments
RT radiotherapy, SIN squamous intraepithelial neoplasia, CIS in situ carcinoma, SCC squamous cell carci-
noma

13
European Archives of Oto-Rhino-Laryngology

Papillomatosis
Yes

Yes

SCC

Perpendicular

SIN3

No
SIN2

No SIN1

Longitudinal

Benign

Previous_treatments ELS_NBI Histology WA_IPCL

Fig. 4  Alluvial chart showing frequency distribution of previous treatment, ELS classification results, histology, and presence of wide angle
IPCL features (WA IPCL). Color code according to different matching of previous treatments and ELS classification results

(67.6%) patients were submitted to endoscopic evalua- 17 (2.4%) received head and neck radiotherapy, and 54
tion and surgical procedures without previous treatments, (7.6%) had been previously biopsied elsewhere. The final
whereas 174 (24.6%) had been already surgically treated, pathologic report was consistent with a benign lesion in 208

Table 3  Agreement analysis by Cohen’s k test

Institution Type of lesion N % Agreement (%) κ 95% CI (κ) p

University of All 434 100 97.7 0.954 0.86–1.0  < 0.0001

Genoa  (Center A) Untreated 284 65.4 97.5 0.949 0.833–1.0  < 0.0001
Previous biopsy/surgery/RT 150 34.6 98.0 0.945 0.785–1.0  < 0.0001
University of All 273 100 94.9 0.872 0.754–0.991  < 0.0001
Brescia  (Center B) Untreated 194 71.1 94.3 0.864 0.723–1.0  < 0.0001
Previous biopsy/surgery/RT 79 28.9 96.2 0.894 0.674–1.0  < 0.0001

13
 European Archives of Oto-Rhino-Laryngology

(29.5%) cases, papillomatosis without atypia in 34 (4.8%), and 0.88, respectively). Considering previous treatments as
mild SIN (SIN1) in 46 (6.5%), moderate SIN (SIN2) in 61 a potential source of bias, for untreated patients the ELS
(8.6%), severe SIN (SIN3) or CIS in 93 (13.1%), and SCC in classification reached the best performance with sensitivity
265 (37.5%). Full details are summarized in Table 2, Fig. 4. and NPV for detection of SCC of 0.93 and 0.95, respectively,
and specificity and PPV for diagnosis of at least SIN1 of
ELS classification interobserver reliability 0.88 and 0.91, respectively. In previously treated patients,
the performance of endoscopic evaluation was still satisfac-
Cohen’s k statistic was used to assess the agreement tory in terms of sensitivity (from 0.82 to 0.86), while it was
between judgment of each lesion by two independent raters poorer in terms of specificity (from 0.34 to 0.49), NPV (from
in each institution applying the ELS classification. Accord- 0.46 to 0.76), and PPV (from 0.45 to 0.84) (Table 4, Fig. 5).
ing to the criteria by Altman et al. [16, 17], reported in The measurement of the CU indexes confirmed this
Table 1, for the entire cohort the result was satisfactory observation with a good CU + Ve and CU−Ve for all out-
and showed a very good agreement between observers at comes except one in untreated patients, whereas no more
both the center A (κ = 0.954; 95% confidence interval [CI] than satisfactory or even poorer results were obtained for
0.86–1.0, p < 0.0001) and the center B (κ = 0.872; 95% CI most outcomes in previously treated or biopsied patients, as
0.754–0.991, p < 0.0001) (Table 3). The agreement was shown in Table 5.
consistent and significant (p < 0.0001) for both Institu-
tions, as well as for untreated (κ = 0.949; 95% CI 0.833–1.0 Diagnostic performance in respiratory
and κ = 0.864; 95% CI 0.723–1.0, respectively) and previ- papillomatosis
ously treated patients (κ = 0.945; 95% CI 0.785–1.0 and
κ = 0.894; 95% CI 0.674–1.0, respectively) (Table 3). Among perpendicular vascular changes, the ELS classifica-
tion well defines the vascular pattern of recurrent respiratory
Diagnostic performance papillomatosis lesions, characterized by vessel loops with
wide angle turning point, embedded in a three-dimensional
Considering the final score in the entire cohort (24 cases warty structure [15]. We tested the identification of these
with initial disagreement were resolved between the exam- features by NBI in our cohort, confirming their value for
iners), performance of the diagnostic test was assessed correct identification of this disease with a sensitivity of 1.0
investigating the detection of at least SIN1-SCC (Table 4). (95% CI 0.90–1.0), specificity of 0.98 (95% CI 0.96–0.99),
The best sensitivity and NPV were obtained for detection PPV of 0.69 (95% CI 0.55–0.82), and NPV of 1.0 (95% CI
of SCC (0.90 and 0.91, respectively) and, accordingly, the 0.99–1.0), as shown in Table 4. Moreover, the measure of
best specificity and PPV for diagnosis of at least SIN1 (0.79 CU indexes confirmed the excellent performance of NBI

Table 4  Diagnostic test results Sensitivity (95% CI) Specificity (95% CI) PPV (95% CI) NPV (95% CI)

All
 At least SIN1 0.80 (0.77–0.84) 0.79 (0.74–0.84) 0.88 (0.85–0.91) 0.68 (0.62–0.73)
 At least SIN2 0.84 (0.80–0.87) 0.75 (0.70–0.80) 0.83 (0.79–0.87) 0.76 (0.71–0.81)
 At least SIN3 0.89 (0.85–0.92) 0.69 (0.64–0.74) 0.75 (0.70–0.79) 0.86 (0.81–0.89)
 SCC 0.90 (0.86–0.94) 0.58 (0.53–0.63) 0.56 (0.52–0.61) 0.91 (0.87–0.94)
Untreated
 At least SIN1 0.80 (0.75–0.84) 0.88 (0.83–0.93) 0.91 (0.87–0.95) 0.74 (0.67–0.79)
 At least SIN2 0.85 (0.80–0.89) 0.86 (0.81–0.90) 0.88 (0.84–0.92) 0.82 (0.76–0.87)
 At least SIN3 0.90 (0.86–0.94) 0.81 (0.75–0.86) 0.81 (0.76–0.86) 0.90 (0.86–0.94)
 SCC 0.93 (0.88–0.96) 0.68 (0.63–0.74) 0.61 (0.55–0.67) 0.95 (0.91–0.97)
Previous biopsy/surgery/RT
 At least SIN1 0.82 (0.75–0.87) 0.49 (0.35–0.63) 0.84 (0.77–0.89) 0.46 (0.33–0.59)
 At least SIN2 0.83 (0.76–0.89) 0.44 (0.33–0.56) 0.75 (0.68–0.82) 0.56 (0.42–0.69)
 At least SIN3 0.85 (0.78–0.91) 0.40 (0.31–0.51) 0.65 (0.58–0.72) 0.68 (0.54–0.79)
 SCC 0.86 (0.77–0.92) 0.34 (0.26–0.43) 0.45 (0.41–0.57) 0.76 (0.63–0.86)
Papillomatosis detection 1.00 (0.90–1.00) 0.98 (0.96–0.99) 0.69 (0.55–0.82) 1.00 (0.99–1.00)

PPV positive predictive value, NPV negative predictive value, CI confidence interval, SIN squamous
intraepithelial neoplasia, SCC squamous cell carcinoma

13
European Archives of Oto-Rhino-Laryngology

Fig. 5  Radar charts showing the diagnostic test applying the ELS surgical or RT treated group (c). Diagnostic test results referred to the
classification for the detection of different histologic targets in the detection of wide angle IPCLs for the diagnosis of laryngeal papil-
whole cohort (a), in the untreated group (b), and in the previously lomatosis (d)

with a CU–Ve of 0.98 (95% CI 0.97–0.99) and good perfor- full integration of high-definition videoendoscopes, easily
mance in terms of CU + Ve of 0.69 (95% CI 0.56–0.83), as activated by pressing a button during in-office endoscopic
reported in Table 5. examination or during pre- and intraoperative assessment.
Interestingly, the superior in-depth evaluation of the bioen-
doscopic features of a given lesion may pave the way to the
Discussion proof of concept of the optical biopsy, i.e. the capability to
understand the nature of a given vocal fold mucosal lesion
Among the several bioendoscopic techniques now avail- before its removal, thus modulating its excisional biopsy and
able for routine evaluation of the UADT, NBI appears to optimizing hospitalization time, costs, and undue damage to
be the most effective in evaluation of the larynx, hypophar- surrounding healthy structures [23].
ynx, oral and oropharyngeal cavities [1, 20, 21]. The easy The need for a common language to categorize and share
use of NBI and other bioendoscopic tools based on similar the findings from NBI evaluation led to a number of different
principles, which aims to enhance the vascular features of classification systems. The first to have widespread diffusion
tissues (e.g. SPIES [22] or iSCAN [20]), is mainly due to in the head and neck scientific community was proposed

13
 European Archives of Oto-Rhino-Laryngology

Table 5  Clinical Utility indexes CU+Ve (95% CI) CU+Ve Judgment CU−Ve (95% CI) CU−Ve Judgment
and utility grading results
according to Mitchell [18] All
 At least SIN1 0.71 (0.67–0.75) Good 0.54 (0.49–0.58) Satisfactory
 At least SIN2 0.70 (0.65–0.74) Good 0.57 (0.53–0.62) Satisfactory
 At least SIN3 0.66 (0.61–0.71) Good 0.59 (0.55–0.63) Satisfactory
 SCC 0.51 (0.45–0.57) Satisfactory 0.53 (0.49–0.57) Satisfactory
Untreated
 At least SIN1 0.73 (0.68–0.78) Good 0.65 (0.60–0.70) Good
 At least SIN2 0.75 (0.69–0.80) Good 0.70 (0.66–0.74) Good
 At least SIN3 0.73 (0.68–0.79) Good 0.73 (0.69–0.77) Good
 SCC 0.57 (0.50–0.64) Satisfactory 0.65 (0.61–0.69) Good
Previous biopsy/surgery/RT
 At least SIN1 0.68 (0.61–0.75) Good 0.23 (0.11–0.34) Poor
 At least SIN2 0.63 (0.55–0.70) Satisfactory 0.25 (0.14–0.35) Poor
 At least SIN3 0.56 (0.47–0.64) Satisfactory 0.27 (0.17–0.38) Poor
 SCC 0.42 (0.32–0.52) Poor 0.26 (0.16–0.36) Poor
Papillomatosis detection 0.69 (0.56–0.83) Good 0.98 (0.97–0.99) Excellent

CU + Ve positive clinical utility index, CU−Ve negative clinical utility index, CI confidence interval, SIN
squamous intraepithelial neoplasia, SCC squamous cell carcinoma

by Ni et al. [11]. These authors divided the different IPCL harboring pre-cancerous or frankly neoplastic alterations.
changes in five types (I–V), judging them as benign (from In particular, we applied this diagnostic tool to demonstrate
types I–IV), suspected malignant, and frankly malignant its possible role in performing a so-called optical biopsy. In
(type V). However, apart from its intrinsic complexity, this fact, our policy has always been, for early glottic lesions, a
classification clearly showed a lack of a clear-cut threshold one-stage modulated excisional biopsy based on a number
between benign and malignant diseases. In fact, different of pre- and intraoperative diagnostic tests in which WL and
authors proposed different cut-offs for the worst endoscopic NBI rigid endoscopy under general anesthesia has always
feature of each lesion to be considered suspicious, ranging played a paramount role [23]. Moreover, as asserted by
from type III [24], to type IV [25–27], and type V [9, 11, many authors, NBI is capable of enhancing small lesions
28]. that are undetectable by WL alone, thus ameliorating the
Therefore, in 2016 the ELS proposed a new classifi- treatment of laryngeal SSC, as well as assessing the poten-
cation system for the interpretation of glottic vascular tial multifocality of the disease and correct evaluation of
abnormalities detected during NBI-guided endoscopies intraoperative margins [31], as well as early identification
[15]. This classification considers vascular abnormalities of small recurrences during follow-up that may still allow
as IPCL perpendicular to the epithelium surface as suspi- application of minimally invasive treatments such as laser
cious, whereas longitudinal vascular changes (e.g. dilated office-based procedures or second-look microlaryngoscopic
or tortuous vessels, increased vessels numbers) are consid- operations [32–35]. Of note, the present study demonstrated
ered as not suspicious to harbor respiratory papillomatosis, a lower diagnostic accuracy of NBI in the previously treated
pre-malignant, or cancerous lesions. The first attempt to patients compared to the untreated ones, thus confirming the
apply this dichotomic classification was in the study by potential confounding factor played by invasive sampling
Šifrer et al. [29] who analyzed 80 vocal cords lesions in procedures when not directed to the full removal (excisional
which the identification of a perpendicular vascular pattern biopsy) of the entire visible lesion within safe margins.
was diagnostic for CIS-SCC with a sensitivity of 100%, The excellent interobserver reliability of the ELS classifi-
specificity of 95%, PPV of 88%, and NPV of 100%. Fur- cation with a k > 0.81 in all scenarios tested and reproducible
ther analysis evaluating a larger cohort of 288 vocal cords in two independent centers confirms the reproducibility of
gave similar results (sensitivity 98%, specificity 95%, PPV the operators’ findings in applying this classification tool.
88%, and NPV 99%) [30]. The high interobserver reliability of the ELS classifica-
Our results, obtained in two of the European pioneer cent- tion can be explained by its intrinsic simple application
ers applying NBI for evaluation of the UADT since 2007, and dichotomic arrangement, providing better performance
herein confirm the intrinsic value of the ELS classification compared to other proposed classification systems such as
for laryngeal vascular changes in the identification of lesions that by Ni, which is complicated by a 5-tier structure and

13
European Archives of Oto-Rhino-Laryngology

associated with moderate/substantial interrater accordance, software applicable in the head and neck, based on artificial
with a k ranging from 0.55 to 0.69 [36, 37]. intelligence algorithms already tested on retrospective stud-
On the other hand, it has to be noted that all the observers ies [45, 46], thus improving the objectivity and detection rate
involved in this study had a minimum experience of 3 years of these diagnostic tools, as already devised for gastrointes-
in the use of NBI technology. Even though application of tinal tract tumors [47, 48].
the ELS Classification on vocal fold vascular changes as
observed by NBI is no more subjective than any other diag-
nostic performance, evaluation of certain subtle and some-
times ambiguous neoangiogenic patterns still may require a Conclusion
higher level of expertise, for which a learning curve is inevi-
tably necessary. However, data derived from the gastrointes- The ELS classification for NBI vascular changes of laryn-
tinal field show that less than a year of training evaluating geal lesions, herein validated in a large multicenter cohort,
200 cases is enough to guarantee an accurate evaluation of is a highly reliable tool with good diagnostic performance in
NBI frames and that the motivation of the trainer itself can the optical biopsy setting, confirming its overall value. The
significantly improve the overall performance [38]. systematic use of this classification seems to allow better
Investigating the diagnostic test, having as a target all the (and purely endoscopic) diagnostic capability of suspicious
possible grades of pre-malignant or malignant transforma- glottic lesions, reliably distinguishing benign ones from
tion, allowed us to depict the capability of the ELS classifi- those with a diagnosis of papillomatosis, SIN, or invasive
cation in helping to correctly identify pre-malignant cases SCC. The excellent performance of NBI for correct identifi-
with the highest PPV and specificity for at least SIN1 diag- cation of respiratory papillomatosis also confirms its useful-
nosis. The lower performance of such parameters observed ness in this clinical setting.
for the final diagnosis of glottic SCC can be explained by
the presence, and progressively increase, of perpendicular
vascular changes at early stages of pre-malignant transfor- Funding  Open access funding provided by Università degli Studi di
Genova within the CRUI-CARE Agreement.
mation (SIN1-SIN2). By contrast, for diagnosis of laryngeal
SCC, the ELS classification had good performance in terms Availability of data and materials  Full dataset will be available at:
of sensitivity and NPV, with a low rate of false negative “ELS_NBI_Classification_Validation_Dataset”, Mendeley Data, V1,
cases and good confidence in a negative result (absence of https​://doi.org/10.17632​/txzzw​9n7xs​.1 https​://doi.org/10.17632​/txzzw​
perpendicular vascular changes). 9n7xs​.1 (Embargo date: 6th December 2021).
Furthermore, several authors have underlined the utility
of NBI for detection of recurrent respiratory papillomatosis Declarations 
and its ability to increase the detection rate of small lesions
Conflict of interest The authors certify that they have no affiliation
that invisible by WL alone [29, 33, 39–42]. The excellent with or involvement in any organization or entity with any financial
performance in terms of CU + Ve and CU−Ve searching for interest.
wide angle IPCLs in the identification of respiratory papil-
lomatosis mandates, as previously suggested by the recent Ethical approval  The research did not involve any animal models; the
research involved human participants in accordance with the ethical
literature [33, 39–44], the use of biologic endoscopy tools standards of the institutional and/or national research committees and
like NBI, and should be considered the endoscopic gold with the 1964 Helsinki Declaration and its later amendments or com-
standard for optical biopsy and follow-up of patients affected parable ethical standards.
by laryngeal papillomas.
Informed consent  Informed consent for disclosure of privacy in man-
The main limits of our study are represented by its ret- aging personal data for scientific purposes was obtained from all par-
rospective design, balanced by analyzing a broad bicentric ticipants included in the study.
cohort. Nevertheless, among the estimator analyzed, the
suboptimal performance in terms of specificity, negative Open Access  This article is licensed under a Creative Commons Attri-
predictive value, and CU−Ve could has been underesti- bution 4.0 International License, which permits use, sharing, adapta-
mated having chosen among the inclusive criteria the need tion, distribution and reproduction in any medium or format, as long
for a histopathological diagnosis: several patients without as you give appropriate credit to the original author(s) and the source,
provide a link to the Creative Commons licence, and indicate if changes
any suspicious lesion at the first evaluation and along time were made. The images or other third party material in this article are
could be considered as true negatives too, thus improving included in the article’s Creative Commons licence, unless indicated
the values of such estimators. otherwise in a credit line to the material. If material is not included in
Further developments in this field might include the anal- the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will
ysis of a prospective cohort of patients, implementing the
enrollment of true negative cases and developing a real-time

13
 European Archives of Oto-Rhino-Laryngology

need to obtain permission directly from the copyright holder. To view a 18. Mitchell AJ (2008) The clinical significance of subjective memory
copy of this licence, visit http://creat​iveco​mmons​.org/licen​ses/by/4.0/. complaints in the diagnosis of mild cognitive impairment and
dementia: a meta-analysis. Int J Geriatr Psychiatry 23:1191–1202.
https​://doi.org/10.1002/gps.2053
19. Mitchell AJ (2011) Sensitivity x PPV is a recognized test called
References the clinical utility index (CUI+). Eur J Epidemiol 26:251–252.
https​://doi.org/10.1007/s1065​4-011-9561-x
20. Hawkshaw MJ, Sataloff JB, Sataloff RT (2013) New concepts
1. Sun C, Han X, Li X et al (2017) Diagnostic performance of nar- in vocal fold imaging: a review. J Voice 27:738–743. https​://doi.
row band imaging for laryngeal cancer: a systematic review and org/10.1016/j.jvoic​e.2013.05.011
meta-analysis. Otolaryngol Head Neck Surg 156:589–597. https​ 21. Deganello A, Paderno A, Morello R et al (2020) Diagnostic accu-
://doi.org/10.1177/01945​99816​68570​1 racy of narrow band imaging in patients with oral lichen planus:
2. Watanabe A, Taniguchi M, Tsujie H et al (2009) The value of a prospective study. Laryngoscope. https​: //doi.org/10.1002/
narrow band imaging for early detection of laryngeal cancer. lary.29035​
Eur Arch Otorhinolaryngol. https ​ : //doi.org/10.1007/s0040​ 22. Staníková L, Walderová R, Jančatová D et al (2018) Compari-
5-008-0835-1 son of narrow band imaging and the Storz Professional Image
3. Chu EA, Kim YJ (2008) Laryngeal cancer: diagnosis and preop- Enhancement System for detection of laryngeal and hypopharyn-
erative work-up. OtolaryngolClin North Am 41:673–695 geal pathologies. Eur Arch Otorhinolaryngol 275:1819–1825.
4. Green B, Cobb ARM, Brennan PA, Hopper C (2014) Optical diag- https​://doi.org/10.1007/s0040​5-018-4987-3
nostic techniques for use in lesions of the head and neck: review 23. Mora F, Carta F, Missale F et  al (2020) Laryngeal mid-cord
of the latest developments. Br J Oral MaxillofacSurg 52:675–680. erythroleukoplakias: how to modulate the transoralCO(2) laser
https​://doi.org/10.1016/j.bjoms​.2014.06.010 excisional biopsy. Cancers (Basel). https​://doi.org/10.3390/cance​
5. Muto M, Nakane M, Katada C et al (2004) Squamous cell carci- rs120​82165​
noma in situ at oropharyngeal and hypopharyngeal mucosal sites. 24. Rzepakowska A, Sielska-Badurek E, Cruz R et al (2018) Narrow
Cancer 101:1375–1381. https​://doi.org/10.1002/cncr.20482​ band imaging versus laryngovideostroboscopy in precancerous
6. Piazza C, Dessouky O, Peretti G et al (2008) Narrow-band imag- and malignant vocal fold lesions. Head Neck 40:927–936. https​
ing: a new tool for evaluation of head and neck squamous cell ://doi.org/10.1002/hed.25047​
carcinomas. Review of the literature. ActaOtorhinolaryngolItalor- 25. Ni XG, Zhu JQ, Zhang QQ et al (2019) Diagnosis of vocal cord
ganoUffdellaSocItal di Otorinolaringol e ChirCerv-facc 28:49–54 leukoplakia: the role of a novel narrow band imaging endo-
7. Piazza C, Del Bon F, Peretti G, Nicolai P (2012) Narrow band scopic classification. Laryngoscope 129:429–434. https​://doi.
imaging in endoscopic evaluation of the larynx. CurrOpinOto- org/10.1002/lary.27346​
laryngol Head Neck Surg 20:472–476 26. Vilaseca I, Valls-Mateus M, Nogués A et al (2017) Usefulness
8. Cohen J (2008) Comprehensive atlas of high resolution endoscopy of office examination with narrow band imaging for the diag-
and narrowband imaging (1st edn) nosis of head and neck squamous cell carcinoma and follow-up
9. Kraft M, Fostiropoulos K, Gurtler N et al (2016) Value of narrow of premalignant lesions. Head Neck 39:1854–1863. https​://doi.
band imaging in the early diagnosis of laryngeal cancer. Head org/10.1002/hed.24849​
Neck 38:15–20. https​://doi.org/10.1002/hed.23838​ 27. De Vito A, Meccariello G, Vicini C (2017) Narrow band imaging
10. Lin YC, Wang WH, Lee KF et al (2012) Value of narrow band as screening test for early detection of laryngeal cancer: a prospec-
imaging endoscopy in early mucosal head and neck cancer. Head tive study. ClinOtolaryngol 42:347–353. https​://doi.org/10.1111/
Neck 34:1574–1579 coa.12728​
11. Ni X-G, He S, Xu Z-G et al (2011) Endoscopic diagnosis of laryn- 28. Rzepakowska A, Sielska-Badurek E, Żurek M et al (2018) Nar-
geal cancer and precancerous lesions by narrow band imaging. J row band imaging for risk stratification of glottic cancer within
LaryngolOtol. https​://doi.org/10.1017/S0022​21511​00020​33 leukoplakia. Head Neck 40:2149–2154. https​://doi.org/10.1002/
12. Takano JH, Yakushiji T, Kamiyama I et al (2010) Detecting hed.25201​
early oral cancer: narrowband imaging system observation of 29. Šifrer R, Rijken JA, Leemans CR et al (2018) Evaluation of vascu-
the oral mucosa microvasculature. Int J Oral MaxillofacSurg lar features of vocal cords proposed by the European Laryngologi-
39:208–213. https​://doi.org/10.1016/j.ijom.2010.01.007 cal Society. Eur Arch Otorhinolaryngol 275:147–151. https​://doi.
13. Barnes L, Eveson JW, Reichart P, Sidransky D (2005) World org/10.1007/s0040​5-017-4791-5
Health Organization Classification of Head and Neck Tumours. 30. Šifrer R, Šereg-Bahar M, Gale N, Hočevar-Boltežar I (2020) The
WHO Classif Tumour diagnostic value of perpendicular vascular patterns of vocal cords
14. Remacle M, Eckel HE, Antonelli A et al (2000) Endoscopic cor- defined by narrow-band imaging. Eur Arch Otorhinolaryngol.
dectomy. A proposal for a classification by the working commit- https​://doi.org/10.1007/s0040​5-020-05864​-5
tee, European Laryngological Society. Eur Arch Otorhinolaryn- 31. Fiz I, Mazzola F, Fiz F et al (2017) Impact of close and positive
gol 257:227–231. https​://doi.org/10.1007/s0040​50050​228 margins in transoral laser microsurgery for Tis-T2 glottic cancer.
15. Arens C, Piazza C, Andrea M et  al (2016) Proposal for a Front Oncol 7:245. https​://doi.org/10.3389/fonc.2017.00245​
descriptive guideline of vascular changes in lesions of the vocal 32. Ivancic R, Iqbal H, deSilva B et al (2018) Current and future man-
folds by the committee on endoscopic laryngeal imaging of the agement of recurrent respiratory papillomatosis. Laryngoscope
European Laryngological Society. Eur Arch Otorhinolaryngol InvestigOtolaryngol 3:22–34. https​://doi.org/10.1002/lio2.132
273:1207–1214. https​://doi.org/10.1007/s0040​5-015-3851-y 33. Ochsner MC, Klein AM (2015) The utility of narrow band imag-
16. Altman DG (1991) Practical statistics for medical research. ing in the treatment of laryngeal papillomatosis in awake patients.
Chapman and Hall. Stat Med. https​://doi.org/10.1002/sim.47801​ J Voice 29:349–351. https​://doi.org/10.1016/j.jvoic​e.2014.08.002
01015​ 34. Zeitels SM, Burns JA, Franco RA et al (2004) Office-based treat-
17. Landis JR, Koch GG (1977) The measurement of observer ment of glottal dysplasia and papillomatosis with the 585-nm
agreement for categorical data. Biometrics. https ​ : //doi. pulsed dye laser and local anesthesia. Ann OtolRhinolLaryngol
org/10.2307/25293​10 113:265–276. https​://doi.org/10.1177/00034​89404​11300​403

13
European Archives of Oto-Rhino-Laryngology

35. Motz KM, Hillel AT (2016) Office-based management of recur- 43. Imaizumi M, Okano W, Tada Y, Omori K (2012) Surgical treat-
rent respiratory papilloma. CurrOtorhinolaryngol Rep 4:90–98. ment of laryngeal papillomatosis using narrow band imaging.
https​://doi.org/10.1007/s4013​6-016-0118-0 Otolaryngol Head Neck Surg (United States) 147:522–524. https​
36. Nogués-Sabaté A, Aviles-Jurado FX, Ruiz-Sevilla L et al (2018) ://doi.org/10.1177/01945​99812​44816​2
Intra and interobserver agreement of narrow band imaging for the 44. Adachi K, Umezaki T, Kiyohara H, Komune S (2015) New tech-
detection of head and neck tumors. Eur Arch Otorhinolaryngol nique for laryngomicrosurgery: narrow band imaging-assisted
275:2349–2354. https​://doi.org/10.1007/s0040​5-018-5063-8 video-laryngomicrosurgery for laryngeal papillomatosis. J Lar-
37. Zwakenberg MA, Dikkers FG, Wedman J et al (2016) Narrow yngolOtol 129:S74–S76. https​://doi.org/10.1017/S0022​21511​
band imaging improves observer reliability in evaluation of upper 40024​36
aerodigestive tract lesions. Laryngoscope 126:2276–2281. https​ 45. Moccia S, De Momi E, Guarnaschelli M et al (2017) Confident
://doi.org/10.1002/lary.26008​ texture-based laryngeal tissue classification for early stage diag-
38. Dias-Silva D, Pimentel-Nunes P, Magalhães J et al (2014) The nosis support. J Med Imaging (Bellingham, Wash) 4:34502. https​
learning curve for narrow-band imaging in the diagnosis of pre- ://doi.org/10.1117/1.JMI.4.3.03450​2
cancerous gastric lesions by using Web-based video. Gastroin- 46. Moccia S, Vanone GO, De ME et al (2018) Learning-based clas-
testEndosc 79:910–920. https:​ //doi.org/10.1016/j.gie.2013.10.020 sification of informative laryngoscopic frames. Comput Meth-
(quiz 983-e1, 983.e4) ods Programs Biomed 158:21–30. https​://doi.org/10.1016/j.
39. TjonPianGi REA, Halmos GB, Van Hemel BM et al (2012) Nar- cmpb.2018.01.030
row band imaging is a new technique in visualization of recurrent 47. Mori Y, Kudo S-E, Misawa M et al (2018) Real-time use of arti-
respiratory papillomatosis. Laryngoscope 122:1826–1830. https​ ficial intelligence in identification of diminutive polyps during
://doi.org/10.1002/lary.23344​ colonoscopy: a prospective study. Ann Intern Med 169:357–366.
40. Lukes P, Zabrodsky M, Lukesova E et al (2014) The role of NBI https​://doi.org/10.7326/M18-0249
HDTV magnifying endoscopy in the prehistologic diagnosis of 48. Guo L, Xiao X, Wu C et al (2020) Real-time automated diagno-
laryngeal papillomatosis and spinocellular cancer. Biomed Res sis of precancerous lesions and early esophageal squamous cell
Int. https​://doi.org/10.1155/2014/28548​6 carcinoma using a deep learning model (with videos). Gastroin-
41. Dippold S, Becker C, Nusseck M et al (2015) Narrow band imag- testEndosc 91:41–51. https​://doi.org/10.1016/j.gie.2019.08.018
ing: a tool for endoscopic examination of patients with laryngeal
papillomatosis. Ann OtolRhinolLaryngol 124:886–892. https​:// Publisher’s Note Springer Nature remains neutral with regard to
doi.org/10.1177/00034​89415​59065​6 jurisdictional claims in published maps and institutional affiliations.
42. Jackowska J, Klimza H, Winiarski P et al (2018) The usefulness
of narrow band imaging in the assessment of laryngeal papil-
lomatosis. PLoS ONE 13:1–9. https​://doi.org/10.1371/journ​
al.pone.02055​54

13