Predictors for postoperative esophageal stricture after balloon-based radiofrequency ablation for early esophageal squamous neoplasia: a multicenter validation study

Abstract

Background:

Endoscopic radiofrequency ablation (RFA) is a rapidly evolving therapeutic modality for early flat esophageal squamous cell neoplasms (ESCNs), but the risk factors for postoperative stricture have not been elucidated. The objective of this study was to identify and validate a predictor for post-RFA stenosis.

Methods:

We consecutively enrolled patients with flat-type ‘large’ (length no less than 3 cm extending no less than half the circumference of the esophagus), early ESCNs, treated with balloon-based RFA (12 J/cm²–clean–12 J/cm² regimen). The tumor and technical factors for postoperative stricture were investigated and we validated the results externally with a society-based multicenter cohort using the same ablation regimen.

Results:

A total of 51 patients were enrolled (30 in the development set and 21 in the validation set). The complete remission rate at 12 months was 93%, and the rates of perforation and postoperative stenosis were 0% and 17%, respectively. Patients with post-RFA stenosis had a significantly larger longitudinal tumor size (mean 115 versus 61 mm, p = 0.003). There were no significant differences in age, body mass index, tumor circumferential extension, pretreatment histological grade, treatment efficacy or size of balloon catheter between the groups with or without stenosis. The optimal cut-off value was set as 9 cm to predict post-RFA stenosis by receiver operating characteristic curve [area under curve (AUC) = 0.881], which was then confirmed to be a reliable predictor by multivariate analysis (odds ratio, 12.7, 95% confidence interval, 1.18–136.28, p = 0.03) and have a good predictive performance in the validation set (AUC = 0.876).

Conclusions:

The most frequent adverse event of RFA was esophageal stenosis, for which the longitudinal tumor size was a significant predictive factor. Early intervention or prevention for stricture should be applied for those with long segment (⩾9 cm) ESCNs.

Introduction

Esophageal cancer is highly lethal and causes more than 400,000 deaths per year worldwide [Ferlay et al. 2010]. With recent advances in image-enhancing techniques such as Lugol’s chromoendoscopy and narrow-band imaging [Hashimoto et al. 2005; Lee et al. 2009, 2010], the number of patients with early esophageal squamous cell neoplasms (ESCNs) detected by local endoscopic treatment has markedly increased. Endoscopic radiofrequency ablation (RFA) is a rapidly evolving therapeutic modality, and recent studies have shown its efficacy and tolerability for eradicating high-grade dysplasia in cases of Barrett’s esophagus [Shaheen et al. 2009, 2011; Phoa et al. 2014], and also for flat-type early ESCNs [Bergman et al. 2011; Van Vilsteren et al. 2011; Wang et al. 2014, 2015; He et al. 2015]. However, postoperative stricture (14–28%) is a major concern with regards to the quality of life after balloon-based RFA, especially when treating squamous neoplasia [Bergman et al. 2011; Van Vilsteren et al. 2011; Wang et al. 2014; He et al. 2015]. Once a patient has started to experience dysphagia due to postoperative stricture, repetitive periodic balloon dilatations are required. However, no systematic study has evaluated the risk factors for adverse events associated with balloon-based RFA, especially esophageal stenosis. In this study, we aimed to evaluate the adverse events associated with esophageal RFA and elucidate the clinical predictive factors for post-RFA stenosis in a prospective derivation cohort and external validation cohort.

Methods

Patients and design

We consecutively recruited adult patients with newly diagnosed, histologically proven early-stage ESCNs (squamous high-grade intraepithelial neoplasms, intramucosal squamous carcinomas) who received balloon-based RFA as the initial treatment at E-Da Hospital, Taiwan, from July 2011 to July 2014. This cohort served as the ‘development set’ to derive the predictors for post-RFA stenosis. All of the enrolled subjects had large, completely flat (type 0–IIb) ESCNs, in which Lugol’s chromoendoscopy showed unstained or mosaic-like lesions occupying more than 50% of the circumference of the esophagus and extending more than 3 cm longitudinally. Endoscopic ultrasound (EUS) revealed no submucosal invasion or lymphadenopathy in any of the tumors, and computed tomography revealed no metastasis or lymphadenopathy. All patients provided informed consent following a full explanation of the use of RFA and alternative treatment options. A complete medical history was obtained before the endoscopic procedures, including demographic and clinical data. The Institutional Review Board of E-Da Hospital approved this study.

Radiofrequency ablation

All of the enrolled patients received balloon-based RFA as the initial treatment. The endoscopic procedures were performed as previously described [Wang et al. 2014]. Briefly, RFA was performed using conscious sedation or anesthesia. Prior to RFA, Lugol’s staining was performed to determine the location and size of the lesions. The area from 1 cm proximal to 1 cm distal to the Lugol-voiding lesion-bearing segment of the esophagus was defined as the treatment area. All primary RFA procedures in the ‘development set’ were performed by a single endoscopist (Dr. W. Wang) using a HALO360 System (Covidien GI Solutions, Sunnyvale, California, USA), that has been approved by the US Food and Drug Administration (FDA) and is approved for use in Europe (CE mark) and Taiwan (Ministry of Health and Welfare). We used a 12 J/cm²–clean–12 J/cm² regimen for all of the procedures [Bergman et al. 2011].

Outcome measures and follow up

After the endoscopic procedures, the patients then received follow-up endoscopy with image-enhancing modalities including Lugol’s staining and narrow-band imaging. This was performed at 1, 3, and 6 months after the procedure and every 6 months thereafter. Two biopsies were taken endoscopically from the normal-appearing mucosa over the treatment area and from Lugol’s unstained areas or any suspicious lesions. If residual squamous intraepithelial neoplasms were detected during follow-up endoscopy, focal ablation with a HALO90 System (12 J/cm², two applications) or Argon plasma coagulation (APC) (Olympus ENDOPLASMA, 1.5 l/min, 35 W) was performed for residual high-grade dysplasia or mucosal cancer. The primary endpoint was the proportion of patients with esophageal stenosis after primary RFA, defined as the failure of a standard endoscope (9.8 mm in diameter) to pass through the stenosis (Figure 1).

Figure 1.

Circumferential balloon-based radiofrequency ablation (RFA) of early squamous neoplasia. (A) Lugol’s staining showed a long-segment Lugol-voiding lesion; (B) Circumferential ablation catheter placed in the esophagus to ablate the lesion; (C) Appearance of the mucosa after the second ablation; (D) Esophageal stricture developed after RFA at 1 month; (E) Balloon dilatation for the stricture; (F) Esophageal stricture resolved after multiple sessions of dilatation, and Lugol’s staining showed no evidence of residual squamous neoplasia.

Statistical analysis

All statistical analyses were performed using SPSS software (SPSS for Windows, version 18.0, SPSS Inc., Chicago, IL). We comparatively analyzed the tumor and technical factors associated with RFA between the patients with and without esophageal stenosis. The tumor factors included location, endoscopic appearance, longitudinal and circumferential tumor size, and histological grade. The technical factors included longitudinal size of ablation, and completion rate after primary ablation. To externally validate the performance of the potential predictors for post-RFA stricture derived from the development set, we collected data from another ESCN cohort of the Taiwan Expert Group of Endoscopic RFA (TEGER), as supported by the Society of Digestive Endoscopy of Taiwan, that was composed of 15 endoscopic experts from 10 medical centers in Taiwan. In this validation cohort, all of the enrolled subjects met the same inclusion criteria, exclusion criteria, and ablation regimens as the development set. Comparisons between the different groups and the clinical characteristics were performed using the χ² test, t-test or Wilcoxon rank-sum non-parametric test as appropriate. The optimal cut-off values of factors to predict post-RFA stenosis were obtained by receiver operating characteristic (ROC) curve analysis. Multivariate logistic regression analysis was performed to determine the independent risk factors for developing esophageal stenosis after RFA. A p value less than 0.05 was considered to indicate statistical significance.

Results

Clinical characteristics and outcomes of the patients

A total of 30 patients with large early flat ESCNs were consecutively enrolled in the development set. The demographics, tumor characteristics and outcomes of these patients are shown in Table 1. The mean longitudinal tumor length was 70 ± 39 mm (range 30–170 mm), and the tumors in 19 of the 30 (63%) patients extended for more than three-quarters of the esophageal circumference. A total of 20 of the 30 (67%) patients achieved complete remission after primary balloon-based RFA, and 28 of the 30 (93%) patients had complete remission at 12 months after additional treatment with focal-type RFA or APC. No perforations occurred during or after the procedure. One patient had a mucosal laceration during the procedure of sizing the esophageal inner diameter, and one had an intramural hematoma that resolved spontaneously without further management. Only 5 of the 30 (17%) patients developed post-RFA stricture that required balloon dilatation to resolve dysphagic symptoms with a median of five sessions (range 2–10). No neoplastic progression or mortality occurred during a mean follow-up period of 17.0 (range 12–48) months.

Table 1.

Clinical characteristics and the outcomes of the patients receiving radiofrequency ablation.

Characteristics	Development cohort (n = 30)	Validation cohort (n = 21)
Age, mean ± SD (range), years	54.0 ± 7.8 (41–76)	56.9 ± 7.9 (46–74)
Male sex	28 (93%)	21 (100%)
BMI, mean ± SD (range), kg/m2	22.0 ± 2.6 (17.0–25.8)	21.1 ± 3.8 (15.1–28.7)
Alcohol drinking	27 (90%)	19 (90%)
Betel-nut chewing	24 (80%)	14 (67%)
Cigarette smoking	25 (83%)	19 (90%)
Tumor location
Upper	9 (30%)	4 (19%)
Middle	13 (43%)	14 (67%)
Lower	8 (27%)	3 (14%)
Tumor longitudinal length, mean ± SD (range), mm	70 ± 39 (30–170)	75.2 ± 38 (30–160)
Circumferential extension of tumor
⩽½	11 (37%)	9 (43%)
⩽¾	19 (63%)	12 (57%)
Histology
HGIN	18 (60%)	12 (57%)
SCC	12 (40%)	9 (43%)
Complete response after primary RFA	20 (67%)	13 (62%)
Complete response at 12 months	28 (93%)	18 (86%)
Adverse events
Perforation	0 (0%)	0 (0%)
Mucosal laceration	1 (3%)	1 (5%)
Intramural hematoma	1 (3%)	0 (0%)
Stricture	5 (17%)	5 (24%)

SD, standard deviation; BMI, body mass index; HGIN, high-grade intra-epithelial neoplasia; SCC, squamous cell carcinoma; RFA, radiofrequency ablation.

Risk factors for postradiofrequency ablation esophageal stricture

Because the most frequently observed adverse event of RFA was esophageal stenosis, we further analyzed the predictive factors for this adverse event in a comparison between the stenosis group (n = 5) and the nonstenosis group (n = 25). The univariate analysis results of the predictive factors for esophageal stenosis are shown in Table 2. The patients with post-RFA stenosis had a significantly larger longitudinal tumor size compared with those without stenosis (mean 115 versus 61 mm, p = 0.003). Those with a tumor located over the upper-third of the esophagus tended to have a higher risk of post-RFA stenosis. There were no significant differences in age, body mass index, tumor circumferential extension, pretreatment histological grade, treatment results or size of balloon catheter between the groups with or without stenosis. Multivariate logistic regression analysis showed that the longitudinal size of the tumors was an independent risk factor for RFA-related stenosis [odds ratio (OR), 1.44, 95% confidence interval (CI), 1.02–2.04, p = 0.04 (Table 3)].

Table 2.

Univariate analysis of predictive factors causing esophageal stenosis after radiofrequency ablation.

Values	With stricture (n = 5)	Without stricture (n = 25)	p value
Age, mean ± SD, years	51.4 ± 6.8	54.5 ± 8.0	0.43
BMI, mean ± SD, kg/m2	21.3 ± 3.6	22.1 ± 2.4	0.50
Concomitant use of ABC	3 (60%)	18 (72%)	0.62
Tumor location at upper esophagus	3 (60%)	6 (24%)	0.14
Tumor longitudinal length, mean ± SD, cm	11.5 ± 3.3	6.1 ± 3.4	0.003
Tumor length > 9 cm	4 (80%)	6 (24%)	0.03
Treatment area, mean ± SD, cm	13.4 ± 3.0	7.8 ± 3.4	0.002
Circumferential extension
⩽½	1 (20%)	10 (40%)	0.62
⩽¾	4 (80%)	15 (60%)
Histology
HGIN	3 (60%)	15 (60%)	0.99
SCC	2 (40%)	10 (40%)
Incomplete ablation after primary RFA	3 (60%)	7 (28%)	0.30
Larger balloon catheter size (⩾25 mm)	1 (20%)	12 (48%)	0.36

SD, standard deviation; ABC, alcohol, betel nut and cigarette; BMI, body mass index; HGIN, high-grade intra-epithelial neoplasia; SCC, squamous cell carcinoma; RFA, radiofrequency ablation.

Table 3.

Multivariate logistic regression analysis to determine the independent risk factor for radiofrequency ablation-related stenosis.

Risk factors	Multivariate analysis
	Adjusted OR	95% CI	p value
Tumor location at upper esophagus	1.40	0.11–17.34	0.79
Tumor longitudinal length, cm	1.44	1.02–2.04	0.04

OR, odds ratio; CI, confidence interval.

Optimal cut-off value and external validation

According to ROC curve analysis, the optimal cut-off value of the longitudinal tumor length was 90 mm to predict post-RFA stenosis with a sensitivity of 80% and specificity of 80% (Figure 2). Multivariate logistic regression analysis revealed that a longitudinal tumor size longer than 90 mm was associated with a significantly higher risk of post-RFA stenosis (OR, 12.7, 95% CI, 1.18–136.28, p = 0.03). We then validated the performance of these factors to predict RFA-related stenosis in a validation cohort from the TEGER (composed of 10 medical centers with 15 endoscopic experts in Taiwan). The demographics, tumor characteristics and outcomes of patients in the validation set are shown in Table 1. Only 5 of 21 patients (24%) developed post-RFA stenosis. ROC curve analysis confirmed the good performance of longitudinal tumor size to predict post-RFA stenosis (the sensitivity, specificity and AUC were 80%, 81.2%, and 88.1%, respectively) (Figure 2). The stenosis rate was not significantly different between the development and validation cohorts (p = 0.722) (Table 4). Of note, 9 of 19 patients with a longitudinal tumor size of more than 90 mm developed post-RFA stenosis, which was significantly more frequent than the patients with a smaller tumor size (47% versus 3%, p = 0.001).

Figure 2.

Receiver-operating characteristic (ROC) curve of tumor longitudinal size to define the optimal cut-off value in the development cohort (Cohort 1) and to validate the performance in predicting postradiofrequency-ablation stricture in the validation cohort (Cohort 2). The areas under the curves were 0.881 and 0.876, respectively.

ROC, receiver-operating characteristic.

Table 4.

Prevalence rate of postradiofrequency ablation stricture for each category of longitudinal tumor size.

Tumor length	3–6 cm	6–9 cm	9–12 cm	>12 cm	p value
Development cohort (n = 30)	0/15 (0%)	1/6 (17%)	2/4 (50%)	2/5 (40%)	0.722*
Validation cohort (n = 21)	0/9 (0%)	0/2 (0%)	3/6 (50%)	2/4 (50%)
Overall (n = 51)	0/24 (0%)	1/8 (13%)	5/10 (50%)	4/9 (44%)	0.001**

^*Comparison between the development and validation cohorts.

^**Comparison between the different categories of tumor size.

Discussion

Endoscopic RFA is a rapidly evolving therapeutic modality for early flat ESCNs; however, the risk factors for developing postoperative stricture have not been elucidated. This study is the first to identify and validate that the longitudinal size of a tumor can be a reliable independent predictor for postoperative stricture, with a cut-off value of 9 cm yielding highest sensitivity (80%) and specificity (80%). Early intervention or prevention strategy for postablation stenosis may be applied for those with long segment (⩾9 cm) early ESCNs.

In this study, we used the same inclusion protocol and the same ablation regimen for all of the enrolled cases, both in the derivation and validation cohorts. We validated the predictive performance in a society-based cohort, and showed that the longitudinal size of the tumor was the key factor for postablation stenosis. In addition, the stricture rates were not different between the single-operator derivation set and the multicenter validation cohort, indicating that the operator was not a key factor for stenosis. With regards to the risk of stenosis, previous studies have compared different regimens of ablation with the use of Lugol’s staining [Scholvinck et al. 2014; He et al. 2015]. Although these studies found that a regimen with single ablation was associated with a lower risk of postoperative stenosis, the 12 J/cm²–clean–12 J/cm² regimen had the highest complete remission rate and is currently the most commonly used regimen, both for Barrett’s esophagus [Shaheen et al. 2009; Van Vilsteren et al. 2013] and ESCNs [Bergman et al. 2011; He et al. 2015]. Single ablation may not be sufficient, especially for squamous carcinomas, because a recent study from the United Kingdom registry demonstrated that 30% of patients with early ESCNs progressed to invasive cancers after only one ablation at each treatment session [Haidry et al. 2013].

To the best of our knowledge, this study is the first to demonstrate that the longitudinal size of a tumor is a reliable factor in predicting stenosis after RFA. In addition, the longer the tumor and the larger the mucosal defect, the higher the risk of contracture. Furthermore, a larger longitudinal lesion led to a higher risk of repetitive ablations or overablation (more than two ablations at the transitional zone). Although the circumferential extension of the tumor has been reported to be a key factor for post-endoscopic submucosal dissection (post-ESD) stricture [Mizuta et al. 2009; Ono et al. 2009; Lee et al. 2012], this study did not disclose the same finding. This may be due to us using ‘balloon-based’ RFA that will ablate the whole circumference of esophagus as the primary treatment for all of the cases, thus the circumferential extension of tumor is not the key factor for post-RFA stenosis. Of note, we also found that tumors located over the upper esophagus tended to develop stenosis. Three of nine patients (33.3%) with a cancer lesion in the upper esophagus were complicated with post-RFA stenosis. In contrast, only 2 of 21 (9.5%) patients with the lesion in the middle or lower esophagus had post-RFA esophageal stenosis. This finding is similar to a previous study with regards to post-ESD stenosis that found that half of the six patients with a cancer lesion in the upper esophagus, and only 11% of 36 patients with the lesion in the lower esophagus were complicated with post-ESD stenosis [Mizuta et al. 2009]. The most likely explanation for this finding lies in differences in the luminal diameter of the esophagus. The diameter of the upper esophagus is smaller than that of the lower esophagus, as could be seen during sizing of the esophageal inner diameter prior to ablation. Furthermore, a previous study related to Barrett’s esophagus with RFA treatment reported that women had a higher risk of postoperative stenosis than men [Pasricha et al. 2015]. Sex differences may be another potential factor for esophageal stenosis. Two female patients in our cohort developed postoperative stenosis, however the sample size was too small to make a definitive conclusion. A larger scale study is needed to clarify the role of sex in post-RFA stenosis.

Esophageal stricture can cause severe dysphagia and result in a decrease in the quality of life and occasionally aspiration pneumonia. Postablation stenosis requires repeated balloon dilatation to resolve the symptoms. For Barrett’s esophagus, a previous study also reported that a long segment length may increase the risk of stricture [Vassiliou et al. 2010]. In this study, we found that 9 of 19 (47%) patients with a longitudinal tumor size of more than 9 cm developed post-RFA stenosis. The high risk of post-RFA stenosis in patients with a tumor size of more than 9 cm suggests that clinicians should routinely perform second-look endoscopy early after RFA, because esophageal stenosis may be evident after a few weeks during the mucosal healing process. Given that there are no current definite predictors of postoperative stenosis for RFA, our findings may help endoscopists when deciding whether or not to perform preventive dilatations for postoperative stricture. Based on our results, prophylactic prednisolone that has been widely used for post-ESD stenosis prevention [Yamaguchi et al. 2011] may be tried in patients with a tumor length of longer than 9 cm.

The tissue effect of RFA on the esophageal wall may explain the mechanism of post-RFA stricture. Most previous studies have used the histology of surgical specimens to evaluate tissue injuries related to RFA in pigs or humans, and showed that the maximal ablation depth (~1000 μm) is the muscularis mucosa layer with an ablation energy of 12 J/cm² in humans [Ganz et al. 2004; Dunkin et al. 2006]. However, this raises the question of why such a thin-layer ablation effect without submucosal damage causes stenosis and should require further study in the future. Although EUS is not sensitive enough to show the efficacy and tissue effect of ablation, RFA, combined with other optic modalities such as optical coherence tomography [Hatta et al. 2012] or confocal laser endomicroscopy, [Guo et al. 2015] with in vivo real-time histological evaluation are potentially practical and valuable.

There are several limitations to this study. First, the sample size was relatively small, and the predictive value of the longitudinal size of a tumor should be confirmed in larger scale prospective studies. Second, the mechanism for post-RFA stenosis is still uncertain. Further studies are needed to elucidate the mechanisms involved, including a possible immunological response and pre-existing fibrous cells in the submucosal layer.

In conclusion, the most frequent adverse event of RFA was esophageal stenosis, for which the longitudinal size of the tumor was a significant predictive factor. Early intervention or prevention should be applied for the patients with a longitudinal tumor size of more than 9 cm.