Blue light imaging and linked color imaging as a screening mode for esophageal squamous cell carcinoma in high‐risk patients: Multicenter randomized trial

Blue light imaging (BLI) and linked color imaging (LCI) are superior to conventional white light imaging for detecting esophageal squamous cell carcinoma (ESCC). Hence, we compared their diagnostic performances in ESCC screening.


INTRODUCTION
2][3] Although Lugol dyeing is the gold standard for ESCC detection, 4-6 it is sometimes painful and time-consuming, resulting in infrequent use of Lugol dyeing for ESCC screening. 7n recent decades, image-enhanced endoscopy (IEE) technologies have been developed and several studies reported their effectiveness in detecting gastrointestinal neoplasms.For instance, a randomized trial revealed that narrow-band imaging (NBI), which can enhance vascular structures, was superior to conventional white light imaging (WLI) in terms of detecting head and neck squamous cell carcinomas (HNSCC) and ESCC. 8In addition, blue light imaging (BLI) has been developed, enabling narrow light observation and good visualization performance similar to NBI. 9,10 Linked color imaging (LCI) is another IEE technology to emphasize color contrast, and some reports have demonstrated its efficacy in detecting gastrointestinal neoplasms. 11,12A recent randomized trial revealed that LCI was more effective than WLI for detecting neoplastic lesions in the upper gastrointestinal tract. 13In subgroup analyses, the superiority of LCI over WLI for the detection of neoplasia was confirmed in the esophagus (2.4% vs. 1.3%) and pharynx (0.9% vs. 0.3%), 13 suggesting that LCI might have the potential to be superior to WLI for the detection of squamous cell carcinoma (SCC).
Both BLI and LCI can be used in the same system.Although these IEEs have been proven to be more effective for the early detection of ESCC than WLI, which is the most optimal screening mode for ESCC remains unclear.Therefore, based on previous results on ESCC detection in each IEE compared with WLI, 8,13 we hypothesized that BLI has a better diagnostic performance than LCI for ESCC screening.To confirm this hypothesis, we conducted this randomized controlled trial.

Study design and ethical issues
T HIS STUDY WAS designed as an open-labeled, parallel, randomized controlled trial at seven hospitals in Japan in accordance with the Declaration of Helsinki, the guidelines of the Consolidated Standards of Reporting Trials, and the Standards for Reporting of Diagnostic Accuracy.The Tohoku Certified Review Board of Tohoku University approved the trial protocol (2020-170), after which this study was registered with the Japan Registry of Clinical Trials (jRCT1022190018-1).Additionally, written informed consent was obtained before eligible patients could participate.

Study populations
We recruited patients at high risk of ESCC as a good cancer screening model.Those with previous ESCC frequently develop metachronous ESCC. 14,15Additionally, patients with HNSCC frequently develop ESCC synchronously and metachronously. 16,17Therefore, we enrolled patients with high risk of ESCC, defined as follows: previous ESCC or HNSCC treated with endoscopic resection or radiation therapy (RT), and present HNSCC.The detail of the eligibility and exclusion criteria are shown in Table S1.

Randomization
Patients were enrolled and randomly assigned to examination groups in a 1:1 ratio by computer-generated randomization using the UMIN-INDICE Cloud System.Randomization was performed by dynamic balancing using the minimization method, with adjustment factors of age (<75 or ≥75 years) and RT history for ESCC or HNSCC.The randomization sequence was concealed from all endoscopists.However, endoscopists and patients were not blinded to the allocated group.
Endoscopists performed esophagogastroduodenoscopy according to the randomization: BLI group (BLI during insertion, followed by LCI during withdrawal) or LCI group (LCI during insertion, followed by BLI during withdrawal).The detail of the procedure is shown in Figure S1.In this study, magnification was not used in any IEE and sedation was used on request.Lugol chromoendoscopy with WLI was performed for the esophagus, as Lugol dyeing is considered the gold standard for ESCC detection, 4-6 and biopsies were performed for any suspected SCCs.Endoscopically suspected SCC was defined as a well-demarcated brownish area or, if visible, irregular microvascular pattern in BLI, a welldemarcated salmon-red to pink area in LCI, and a welldemarcated unstained area in the Lugol chromoendoscopy. 6,7,13,18,19During each procedure, endoscopists recorded whether any suspicious lesions were detected, the confidence in the probability of SCC (high or low), and the details of any detected lesions.If the lesions were detected by BLI/LCI but unrecognized after the Lugol chromoendoscopy due to iodine staining, re-examination was performed to take biopsies.Their Lugol-voiding lesion (LVL) grades were recorded according to a previous report. 20To maintain the endoscopic quality in this study, all participating endoscopists were expert endoscopists certified by the Japan Gastroenterological Endoscopy Society or those with experience in over 2000 cases of esophagogastroduodenoscopy and for more than 5 years.All participating endoscopists were trained using representative BLI and LCI images of superficial ESCC before the start of this trial.
diagnosis was completed after central review by an expert pathologist (F.F.), who was blinded to the allocated group and recorded endoscopic findings.When ESCC was suspected in the Lugol chromoendoscopy findings, such as a pink-color sign, re-examination or endoscopic resection was performed even if the case was not diagnosed as SCC in the biopsy specimen.A case with diagnostic discrepancies between the biopsy and resected specimen was finally diagnosed on the basis of pathology observed in the resected specimens.

Study end-points
The primary end-point was the detection rate of ESCC in the primary mode.The secondary end-points were as follows: (i) a miss rate of ESCC in the primary mode; (ii) diagnostic parameters in the primary mode; and (iii) the detection and miss rates of pharyngeal/laryngeal SCC in the primary mode.Although this study included patients with present pharyngeal/laryngeal SCC, we evaluated only concomitant pharyngeal/laryngeal SCC or ESCC in such cases.A missed lesion was defined as that detected by the secondary mode or Lugol chromoendoscopy but not by the primary mode, and the miss rate was calculated by dividing patients with missed lesions by the total number of patients with at least one ESCC detected.Diagnostic parameters were calculated in a perlesion analysis.Subgroup analyses for ESCC were conducted based on the RT history for the ESCC, LVL grades, sex, sedation, and type of endoscope.Those for pharyngeal/ laryngeal SCC were also conducted based on the sedation situation and type of endoscope.Then, procedure time was measured from the start to detect suspicious lesions.

Sample size
The sample size was calculated based on previous reports about patients with HNSCC and ESCC treatments.In a meta-analysis, 15% of the patients with HNSCC were diagnosed with esophageal secondary primary tumors by Lugol chromoendoscopy. 16The detection rate of metachronous ESCC after endoscopic resection by Lugol chromoendoscopy was approximately 5%. 22While the diagnostic performance of NBI has been considered comparable to Lugol chromoendoscopy, 23 BLI is considered to have similar diagnostic ability for ESCC to NBI. 24 Expecting the ratio of HNSCC patients and follow-up patients after the treatment for ESCC as 1:1, the detection rate of ESCC by BLI was estimated to be 10%.Although no data are available regarding the detection rate of ESCC by LCI, the detection rate of ESCC by NBI was three times higher than that by WLI, 8 and that by LCI was 1.6 times higher than that by WLI. 13 Therefore, the detection rate of ESCC by LCI was assumed to be 5%.Finally, the sample size for detecting significant differences was set at 700 patients to achieve a two-sided alpha of 0.05 with a power of 0.80.

Statistical analyses
Our analysis was performed on an intent-to-treat basis, defined as all randomized patients with informed consent.We counted excluded cases after randomization as not having any neoplastic lesion.Pearson's v 2 -test or, if appropriate, Fisher's exact test was used to compare proportions.The other continuous data between the two groups were compared using the Mann-Whitney U-test.Differences were expressed as risk ratio (RR) with a 95% confidence interval (CI).Additionally, we calculated RRs for the detection and miss rates of ESCC by each subgroup and used tests of interaction (the Mantel-Haenszel test) to identify significant differences.Differences were considered significant if the P-value was below 0.05.An independent statistician (T.N.), using R version 4.2.1 (The R Foundation, Vienna, Austria), performed all the statistical analyses.

RESULTS
Patient enrollment and group characteristics B ETWEEN JANUARY 2020 and April 2021, 700 patients were enrolled in this study.A flow diagram of patient enrollment and randomization is presented in Figure 1.The baseline characteristics in the two groups were well balanced (Table 1).The assessment of procedure times showed no significance between the two groups (Table 2), and no adverse events occurred during these procedures.The representative endoscopic images of ESCC, background esophagus, and pharyngeal/laryngeal SCC are shown in Figure 2 and Figure S2.The pathological results in the two groups are shown in Table 2.

Detection and miss of ESCC
The detection rate of ESCC in the primary mode was 4.0% (14/351) in the BLI group and 4.9% (17/348) in the LCI group, with no significant difference (RR 0.82; 95% CI 0.41-1.63)(Table 3).However, the number of patients with ESCC, as diagnosed by Lugol chromoendoscopy, followed by biopsy or resection of the Lugol-unstained region in the BLI group (19 patients with 20 lesions) tended to be smaller than that in the LCI group (30   S2).In the analyses limited to male individuals, results similar to those of the whole patients were also obtained (Table S3).
The results from the subgroup interaction test are presented in Table 4, wherein the detection of ESCC in the primary mode of the BLI or LCI groups had no significant interaction across the subgroups.Similar results were acquired in miss of ESCC.

Detection and miss of pharyngeal/laryngeal SCC
Overall, pharyngeal/laryngeal SCC was detected in 2.0% (7/ 351) and 0.9% (3/348) of the BLI and LCI groups, respectively.The detection rate of pharyngeal/laryngeal SCC in the primary mode had a tendency to be higher in the BLI group than in the LCI group (2.0% [7/351] vs. 0.3% [1/ 348]; P = 0.069) (Table 3).The miss rate of pharyngeal/ laryngeal SCC in the primary mode tended to be lower in the BLI group (0.0% [0/7] vs. 66.7% [2/3]; P = 0.067).The results of the subgroup analyses according to the sedation situation and type of endoscope are presented in Table S4.

Diagnostic parameters
Table 5 provides the diagnostic parameters of BLI and LCI examined as the primary mode in ESCC and pharyngeal/ laryngeal SCC.In ESCC, while the sensitivity of BLI was significantly higher than that of LCI (75.0% vs. 47.6%;P = 0.042), the positive predictive value (PPV) of BLI tended to be lower than that of LCI (28.8% vs. 45.5%;P = 0.092).When SCC/HGIN was targeted, similar results were obtained (Table S5).

DISCUSSION
F OR THE CURATIVE treatment of ESCC, early detection is deemed mandatory.Based on two randomized trials, 8,13 NBI, which enables narrow light observation similar to BLI, 24 and LCI were found to be more effective than WLI for detecting neoplastic lesions in the esophagus.However, it remains to be elucidated which of BLI or LCI is optimal as a screening mode for ESCC.Thus, we conducted this randomized trial.
Contrary to our hypothesis, the detection rate of ESCC by the primary mode, which was set as a primary end-point, did not significantly differ between the BLI and LCI groups (4.0% vs. 4.9%).The main causes of the unmet study would be a lower overall detection rate of ESCC, especially in the BLI group, than the estimated rate (5.4% vs. 10.0%) and lower sensitivity of BLI than expected (75% vs. 100%), which led to an insufficient sample power.This insufficiency could have caused unmet results because of the accidental imbalanced prevalence of ESCC between the BLI and LCI groups (19 vs. 30 patients).
As shown in the previous studies, a difference in the miss rate is more likely to reach statistical significance than the difference of detection rate. 25This study revealed a reduced miss rate of ESCC in BLI compared with LCI.Interestingly, while no cases with ESCC missed by BLI were detected by LCI, 78.9% of cases with ESCC missed by LCI were detected by BLI.Furthermore, the sensitivity was significantly higher in BLI (75.0% vs. 47.6%), which is important to reduce the false negative leading to the fatal result for missing cancer.These results would have been much less affected by the imbalanced prevalence of ESCC between the two groups than the detection of ESCC in the primary mode because the outcomes were calculated based on the patients with ESCC, but not all included patients.
However, the lower tendency of PPV in BLI could not be ignored.The PPV of BLI in this study (32.7% for ESCC or HGIN) was relatively similar to that of NBI  for ESCC or HGIN in previous studies (34.3-36.1%),and thinning of the keratinous layer and the thinning of epithelium, and an extravascular component of hemoglobin were reported as significant factors for brownish epithelium in NBI. 26,27Nonneoplastic regions with such factors may have caused the low PPV in BLI  or NBI.Based on previous studies, 7,23,28 BLI with magnification, but not Lugol chromoendoscopy, may contribute to improving the PPV.Furthermore, the artificial intelligence system showed high PPV for ESCC, 29 which might overcome this issue.The procedure time for the esophagus in this study was 29-34 s, which may be longer than that in daily practice; however, this time was a little shorter than that in previous reports (38-63 s). 13,30Although, to our knowledge, no studies have evaluated the appropriate observation time limited to the whole esophagus in screening endoscopy, total examination time of esophagogastroduodenoscopy >5 min may improve the detection of esophageal lesions. 31his study also showed a higher tendency of detection and a lower tendency of missed pharyngeal/laryngeal SCC in BLI.However, it should be noted that, in addition to a very small number of cases with pharyngeal/laryngeal SCC, the sedation situation and type of endoscope, which would largely affect the diagnostic performance of pharyngeal/ laryngeal SCC, [32][33][34][35] was not strictly controlled in this study.Although this study was planned mainly for revealing the detection and miss of ESCC, a very large-scale study with a strict control of sedation situation and type of endoscope is required to make a definite conclusion about pharyngeal/ laryngeal SCC.
Several strengths of this study warrant mention.First, the study design was a prospective, multi-institutional randomized style.Second, the real-time on-site diagnoses during the examination reflect the real clinical situations.Finally, unlike prior randomized trials, 8,13 we used Lugol chromoendoscopy to evaluate ESCC in all patients, enabling us to reduce overlooked ESCC.
This study has several limitations.First, insufficient sample power may have led to unmet results in the primary outcome, as mentioned above.Second, observation bias cannot be excluded, since the endoscopists were not blinded to the allocated group.Third, it remains uncertain whether similar results could be observed in patients with a low risk of ESCC.Fourth, the sedation situation and type of endoscope were not strictly controlled in this study.3][34][35] Lastly, we did not unify the air insufflation degree, which was proven to be an important factor to visualize ESCC. 36n conclusion, BLI was not proven to be superior to LCI in the detection rate of ESCC.Meanwhile, BLI was superior to LCI in reducing the miss rate and high sensitivity for ESCC, suggesting that BLI may have the potential to be advantageous over LCI for the diagnosis of ESCC.However, it is still unclear whether BLI is superior to LCI in the diagnostic performance of ESCC and a further largescale study is needed.

Figure 1
Figure 1 CONSORT flow diagram showing the patient enrollment process.BLI, blue light imaging; LCI, linked color imaging.

Table 1
Baseline characteristics of the study patients †One patient in the blue light imaging (BLI) group and seven in the linked color imaging (LCI) group had a history of radiation therapy (RT) for both esophageal squamous cell carcinoma (ESCC) and head and neck squamous cell carcinoma (HNSCC).‡One patient in the BLI group was excluded due to the cancellation of the examination.§ This grade was recorded according to the number of Lugol-voiding lesions (LVLs) per endoscopic view: A, no lesions; B, 1-9 lesions; C, ≥10 lesions.¶ Three patients in the BLI group and one in the LCI group were excluded because Lugol staining was not performed.IQR, interquartile range.

Table 2
Pathological results and procedure time to detect in the blue light imaging (BLI) and linked color imaging (LCI) groups BLI group LCI group P-value ¶Time to detect suspicious lesions in esophagus and pharynx/larynx.HGIN, high-grade intraepithelial neoplasia; IQR, interquartile range;LGIN, low-grade intraepithelial neoplasia.

Table 3
Detection Nineteen patients with 20 ESCC lesions in the BLI group and 30 patients with 42 ESCC lesions in the LCI group.‡Six patients, who had both lesions detected and missed by the primary mode, in the LCI group were counted as missed patients as well as detected patients.§The miss rate was calculated by dividing patients with missed lesions by the total number of patients with at least one ESCC detected.CI, confidence interval.
and miss rates of esophageal squamous cell carcinoma (ESCC) and pharyngeal/laryngeal squamous cell carcinoma (SCC) in the blue light imaging (BLI) and linked color imaging (LCI) groups

Table 4
Subgroup interaction tests showing the detection and miss of esophageal squamous cell carcinoma One patient in the BLI group was excluded due to the cancellation of the examination.¶ Nine patients who underwent the examination by thin endoscopy were under a sedated condition.CI, confidence interval; RT, radiation therapy.
†This grade was recorded according to the number of Lugol-voiding lesions (LVLs) per endoscopic view: A, no lesions; B, 1-9 lesions; C, ≥10 lesions.‡Three patients in the blue light imaging (BLI) group and one in the linked color imaging (LCI) group were excluded because Lugol staining was not performed.§