FLEX October 2023

15314995, 2017, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/lary.26263 by Karuna Dewan - Ochsner Medical Foundation - New Orleans - USA , Wiley Online Library on [21/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

When examining patients, clinicians are capable of seeing the whole video, taking a closer look at the lesi ons, and have the option to alternate between WL and NBI. Nonetheless, in our study design, we wished to par ticularly distinguish the value of NBI from WL. This necessitated a highly comparable presentation for the two imaging modalities, which was difficult to achieve with video; thus, the still image presentation was cho sen. Moreover, alteration of the two imaging modalities had to be omitted. Therefore, we believe that although the true sensitivity and specificity rates may be higher than those observed in our study, the relationship of sen sitivity and specificity between WL and NBI reported in our study is highly reliable. In contrast to previous studies, our study relied on the interpretation of six different independent observers. Although we noticed substantial interobserver differ ences in absolute lesion size, we found a similar trend in five of the six observers with a high ICC regarding in creased lesion size using NBI. Also, five of the six observers tended to estimate more lesions as “invasive carcinoma” by NBI. This implies that the interpretation tendencies found in our study are common to different types of observers. There are several limitations to the office NBI. The quality of the flexible optic image is inferior to that of the rigid telescope in terms of brightness, contrast, and resolu tion. The awake examination is influenced by patient coop eration and conditions. Images may be obscured by mucous, swallowing, coughing, and gagging. These may also cause temporary diffuse vascular congestion that changes the overall appearance of the tissue. The intraoperative images are taken from a distance of a few millimeters under relaxa tion and suction. Such conditions facilitate easy identifica tion of IPCL classes. This is more difficult to achieve under a flexible endoscope office examination. Nevertheless, the office examination is the key point from which the surgeon will make the decision whether to go to the OR. The results of our study suggest that despite its relative inferiority, pre operative flexible NBI may serve to improve evaluation of the extent of suspected lesions, along with a higher detec tion rate of malignancy. Apart from preoperative evaluation, numerous stud ies have been published regarding the utility of NBI along with other stages in the pathway of managing on cologic patients. Pan et al. used NBI to guide office transnasal biopsies from patients unable to undergo gen eral anesthesia due to difficult airways. 19 During direct laryngoscopy, NBI may increase the accuracy of biopsy taking while adhering to IPCL classification. 20,21 Garo folo et al. 15 found that routine use of NBI with a rigid telescope and HDTV during TLM reduced the percent age of positive superficial margins to only 3.6%, com pared with a historical control cohort without NBI having 23.7% positive margins. During the postoperative period, Irjala et al. 13 described three cases in which NBI findings in follow-up for carcinoma or papilloma led to a biopsy and consequent diagnosis of carcinoma or dyspla sia. Postirradiation or postchemoradiation follow-up is challenging due to diffuse local changes in the mucosal surfaces. Piazza et al. 14 studied 12 patients and showed

TABLE III. Performance Measurements for Identification of Carcinoma by NBI and WL.

Performance Measurement

WL (95%CI)

NBI (95% CI)

Sensitivity

48.7% (39.4%–58.1%) 76.1% (68.1%–82.9%) 63.3% (52.5%–73.2%) 63.6% (55.8%–71.0%)

58.62% (49.1%–67.7%) 61.19% (52.4–69.5%) 56.7% (47.3%–65.7%) 63.1% (54.2%–71.4%)

Specificity

PPV

NPV

Positive is defined as carcinoma diagnosis, negative as dysplasia, papilloma, keratosis, or benign. CI 5 confidence interval; NBI 5 narrow band imaging; NPV 5 nega tive predictive value; PPV 5 positive predictive value; WL 5 white light.

Advantages of office NBI examination are many. It is a built-in feature embedded in the endoscopic tool. Its utili zation does not require a separate examination, but merely a button to switch between WL and NBI modalities. This facilitates alternating comparison between WL and NBI. Watanabe et al. 16 examined 35 cases for suspected malig nancy and achieved 91.3% and 91.6% specificity with office NBI examination when compared to final pathologic reports. Qi et al. 17 examined a much larger series com posed of 1,153 case, most of them benign. NBI predicted 166 of 168 malignant cases, with only two false negatives. Kraft et al. 18 evaluated 205 cases (of which 57 were malig nant) with WL or WL 1 NBI, using the Ni classification 11 to differentiate malignant (IPCL type V) from nonmalignant (IPCL types I–IV). They achieved 97% sensitivity and 96% specificity. In those studies, separate interpretation of NBI by independent observers was not utilized; for instance, in the latter study three experts viewed the endoscopic videos together while alternating between NBI and WL. The sen sitivity rate for preoperative flexible NBI in our study resembled the rate reported by Piazza et al. 12 : 59% and 61%, respectively; however, with much lower specificity: 61% compared with 87%. Piazza’s methods also included joint evaluation by three experts. The differences between the sensitivity and speci ficity in our study and those reported by previous stud ies may have several explanations: 1. Our observers analyzed still images and not video. 2. Images were relatively far from the lesion to encompass the whole glottis. Delicate IPCL structures may not have been detected. Unlike previous studies, one of our goals was to assess size differences between WL and NBI, a subject not yet having been addressed. Therefore, images selected in our study demonstrated the glottis completely. 3. Our observers could see only one imaging modality at a time and could not compare WL to NBI for each case. 4. All of our cases were suspected of malignancy. Examiners had to choose the probable diagnosis from seven multiple choice options. Observers may subconsciously have tended to spread their answers over all seven options, although the real distribution of lesion pathology in the study group did not act this way. 5. The observers were independent, and no consultation was allowed.

Laryngoscope 127: April 2017

Shoffel-Havakuk et al.: NBI Preoperative Detection of Malignancy

898