Is Lugol chromoendoscopy omissible in screening for esophageal squamous cell carcinoma?

The development of multiple metachronous cancers in the remaining esophagus after treatment of esophageal squamous cell carcinoma (ESCC) is a long-term management problem. The annual incidence of multiple metachronous cancers, especially after endoscopic resection, is approximately 10% per year.¹ In particular, the grade according to the number of Lugol-voiding lesions (LVLs) per endoscopic view has been reported to stratify the risk.¹ Lugol chromoendoscopy (LCE) is useful for the detection of ESCC and is the standard practice worldwide. However, with the advancement of image-enhanced endoscopy (IEE), narrow-band imaging (NBI) has proven to be significantly superior to white-light imaging (WLI) in detecting ESCC² and is widely used in routine clinical practice. Blue-light imaging (BLI) is considered a modality similar to NBI in the esophagus. As the image quality of IEE has improved, the possibility of replacing LCE with NBI and BLI has been discussed. To date, a meta-analysis using expert-focused studies has shown that NBI has a sensitivity comparable to that of LCE but superior specificity in identifying high-grade dysplasia and/or squamous cell carcinoma of the esophagus.³ In addition, as a prospective randomized controlled trial that included nonexperts revealed the higher specificity of NBI, the latest European guidelines recommend narrow light observation as an alternative to LCE for detecting ESCC.⁴ However, additional LCE was proposed to improve the detection of synchronous lesions, which were easily missed by the nonexperts in this trial; thus, the positioning of LCE and NBI is controversial.

In this issue of Digestive Endoscopy, Ogata et al.⁵ conducted a post hoc analysis of a multicenter randomized controlled trial that compared the diagnostic ability of BLI and linked-color imaging for ESCC in patients at high risk for ESCC in expert settings. In this study, the authors evaluated the diagnostic abilities of BLI and LCE after BLI. Finally, BLI had a significantly lower sensitivity (83.7% vs. 100.0%) and higher specificity (88.2% vs. 81.2%) and accuracy (87.8% vs. 82.5%) than LCE following BLI in the per-patient analysis. In contrast, BLI had a slightly lower detection rate (5.9% vs. 7.0%) and a significantly higher miss rate (18.4% vs. 4.1%) than LCE following BLI. Because the order of examination is usually IEE observation followed by chromoendoscopy, it is difficult to compare the true diagnostic abilities of digital imaging and chromoendoscopy. The strength of this study may be its characterization of BLI and LCE in the diagnosis of superficial ESCC and its suggestion that these may complement each other. Ogata et al. analyzed the characteristics of lesions missed using BLI. Although flat lesions tended to be more frequent in missed lesions than in detected lesions, all lesions missed by BLI were confined to the epithelium/lamina propria mucosa (pT1a-EP/LPM). Other reports similarly reported that many of the lesions missed by NBI/BLI were flat^{6, 7} and most were pT1a-EP/LPM lesions, but some were confined to the muscularis mucosae or submucosa.⁶ The results by LVL grade showed that the proportion of LVL grade A was low for both BLI-detected and BLI-missed lesions (2.4% and 0.0%, respectively), whereas the proportion of grade C was higher (63.4% and 88.9%, respectively). These results may provide insights into the use of BLI and LCE, such as omitting LCE in LVL grade A cases.

Subsequently, Ogata et al. attempted to logically explain their results by objectively comparing BLI and LCE using color difference analysis between cancerous and noncancerous regions. This showed a higher color difference in LCE than in BLI, which supports the results of higher sensitivity in LCE following BLI and is compatible with our impression during examinations.

Furthermore, 19 of the 50 lesions detected in this study had a history of radiotherapy for ESCC, 16 of which were detected by BLI and the remainder were missed by BLI. Chemoradiotherapy (CRT) for ESCC has become widely used as a curative strategy, and the importance of surveillance after CRT for ESCC is increasing. However, differential diagnosis with local residues, recurrence, metachronous ESCC, and usual post-CRT sequelae such as mucositis or fibrosis remains an unmet issue in the endoscopic observation for patients after CRT. This is because unspecified LVLs similar to ESCC can be shown in the esophageal mucosa after CRT. Asada-Hirayama et al. reported that NBI with magnification showed equal sensitivity and significantly higher PPV (85.7% vs. 8.3%) compared to LCE in detecting ESCC or high-grade intraepithelial neoplasia in the patients with a history of CRT.⁸ Therefore, IEE may have a potential to detect metachronous ESCC efficiently in patients after CRT.

In recent years, artificial intelligence (AI) has been developed for endoscopy. Using AI assistance in diagnosing of ESCC has improved the diagnostic accuracy of both novice and expert endoscopists. Additionally, in a large multicenter, tandem randomized controlled trial of more than 10,000 cases, it was reported that the AI-first group reduced, but without reaching significance, the per-lesion miss rate of ESCC and precancerous lesions (1.7% vs. 6.7%) and improved the detection rate (1.8% vs. 1.3%) compared with the routine-first group.⁹ The study involved endoscopists from different types of institutions with a variety of experience and skills; therefore, the AI system is expected to play a role as a valuable auxiliary tool to facilitate the homogenization of ESCC diagnosis for endoscopists at all levels. Because this study specified observations using WLI and NBI, and LCE was used in only approximately 2% of cases, the addition of LCE was not analyzed. On the other hand, AI systems using WLI and NBI imaging that predict multiple LVLs have also been reported. Future research may lead to the development of AI technology that surpasses LCE.

Although LCE is known to be a reliable method for the detection of ESCC, it is problematic in terms of safety and symptoms such as heartburn, chest pain, and nausea caused by the sprayed potassium iodide solution, as well as a significantly longer examination time. Therefore, attempts have been made to omit LCE. Although studies within a single endoscopy session often focus on the diagnostic ability of the lesion and how much is missed, these studies are strongly associated with surveillance methods and the true benefit to the patient is debatable. If the surveillance interval can be shortened, it may not matter if some cEP/LPM lesions are missed; however, if the surveillance interval is long, the number of missed cancers should be minimized. Of course, lesions deeper than cMM/SM1 (tumor invasion of the muscularis mucosae or submucosa ≤200 μm) cannot be missed regardless of the surveillance interval, and if they are missed, the risk of lymph node metastasis is a concern and may affect the quality of life and prognosis of the patient. Current Japanese guidelines recommend endoscopy at least once per year,¹⁰ whereas endoscopy every 3 to 6 months is recommended in Europe,⁴ and there are currently no uniform recommendations for surveillance. Furthermore, surveillance methods should be considered according to the risk of multiple metachronous ESCC, such as the LVL grade. Thus, a prospective randomized trial that includes surveillance methods and surveillance intervals, taking LVL grade into account, is desirable.

This study provides useful insights into routine practice in this field and raises interesting questions regarding ESCC detection. It is also important to understand the weaknesses of BLI and NBI observations and decide on the observation method on a case-by-case basis, depending on the risk of multiple metachronous ESCC and the surveillance interval. Further studies are required to determine appropriate surveillance methods.

Author T.Y. is an Associate Editor of Digestive Endoscopy and T.Y. receives research grants from Fujifilm and Olympus outside of this work. T.K. declares no conflict of interest for this article.

None.