Risk factors of postoperative stricture after endoscopic submucosal dissection for superficial esophageal neoplasms

Abstract

Background:

As larger-sized superficial esophageal neoplasms became candidates for endoscopic submucosal dissection (ESD), post-ESD esophageal stricture has inevitably developed into a significant complication during long-term follow-up.

Method:

The PubMed, Cochrane Central Register of Controlled Trials, MEDLINE, EMBASE, Web of Science, as well as China National Knowledge Infrastructure, the Wanfang Database, and the Chinese Biomedical Literature Database, were searched to identify all the appropriate studies published from January 2000 through October 2019. For risk factor assessment between postoperative stricture and control groups, pooled odds ratios (OR) and weighted mean differences (WMD) estimation was done. All meta-analytical procedures were conducted by using Stata version 15.1 software.

Results:

The results showed that 11 studies with 2248 patients (284 structure cases and 1964 controls) were eligible for this meta-analysis. Statistical results indicated 6 substantial risk factors: lesion characteristics involving the upper third of the esophagus (OR 1.51, [1.02–2.25]), macroscopic type of IIa/IIc (OR 2.76, [1.55–4.92]), tumor depth of invasion above m1 (OR 7.47, [3.31–16.86]), and m2 (OR 12.67, [4.00–40.10]), longitudinal length (WMD 13.75 mm, [7.76–19.74]), circumferential diameter (WMD 10.87 mm, [8.13–13.60]), and circumferential range >3/4 (OR 38.17, [9.94–146.52]). Each additional 10% of the circumferential range increased the risk of stricture by 149% (OR 9282.46, [978.14–88089.35]).

Conclusions:

Six risk factors were assessed to have a key role in the elevated risk levels of post-ESD esophageal stricture. The results can help doctors identify patients with increased risk and thus can guide management of the adequate period of surveillance after ESD and take available approaches of stricture prevention.

1. Introduction

China is a country with a high incidence of esophageal neoplasms in the world. The incidence was 20.35/100,000, accounting for 46.6% of the world, and the mortality was 15.17/100,000, ranking 6th and 4th in the incidence and cause of death of malignant tumors respectively, accounting for 7.52% and 9.26% of all malignant tumors respectively.^[1] The early diagnosis rate of esophageal neoplasms was only 1.43%. More than 90% of patients have progressed to the middle and late stages when clinically diagnosed. The overall 5-year survival rate was <20%, and the postoperative 5-year survival rate was hovering around 30%. The 5-year survival rate for early esophageal neoplasms was >95%.^[2]

With the development of endoscopy techniques and instruments, more superficial esophageal neoplasms can be detected early. Endoscopic submucosal dissection (ESD) has been considered as an effective and minimally invasive treatment for superficial esophageal neoplasms than esophagectomy.^[3–11] ESD is more effective in comparison to endoscopic mucosal resection (EMR), in achieving curative resection, en-bloc resection, precise histopathologic assessment, and prevents esophageal neoplasia from occurring again.^[12–14]

As larger-sized superficial esophageal neoplasms became candidates for ESD, post-ESD esophageal stricture has inevitably become a serious problem during long-term follow-up.^[15–18] Patients suffering from post-ESD esophageal stricture have symptoms of nausea, vomiting, and varying degrees of dysphagia, resulting in decreased quality of life.

The risk of postoperative stricture is influenced by all factors of the patient, features of the lesion, particulars of the procedure, and postoperative. Over time, various studies have highlighted this problem to identify the risk factors. It has been reported that circumferential range is an important predictor of postoperative stricture. If the extent of circumferential resection is >50% of esophageal circumference, the rate of stricture occurrence can reach 6.9% to 18%^[19–21] and increases to 70% to 80% if the extent is over 60%.^[22] In patients undergoing resection of near-circumference or whole-circumference, the risk of stricture is almost 100%, if the resection area comprises a long esophageal segment.^[4,17,23,24]

Research is still lacking to ensure the risk factors of postoperative stricture for other aspects. Therefore, this study is intended to critically review the published data and analyze the effects of patient factors, lesion characteristics, procedural details, and postoperative characteristics on the occurrence of postoperative stricture.

2. Materials and methods

2.1. Search strategy

The design development, research conduction, and reporting of this meta-analysis were carried out by following the statements of PRISMA and MOOSE.^[25,26] The PubMed, MEDLINE, EMBASE, Web of Science, Cochrane Central Register of Controlled Trials, as well as China National Knowledge Infrastructure, the Wanfang Database, and the Chinese Biomedical Literature Database, were searched to identify all the appropriate studies published from January 2000 through October 2019, with the following search strategy: (endoscopic submucosal dissection OR ESD) AND (esophagus cancer OR esophagus cancers OR esophagus neoplasm OR esophagus neoplasms OR esophageal cancer OR esophageal cancers OR esophageal neoplasm OR esophageal neoplasms OR cancer of esophagus OR cancer of the esophagus OR superficial squamous esophageal neoplasms OR superficial esophageal cancers OR esophagus squamous cell carcinoma OR esophagus adenocarcinoma) AND (stricture OR stenosis). This meta-analysis on risk factors of postoperative stricture after endoscopic submucosal dissection for superficial esophageal neoplasms was conducted on the basis of published articles, and no privacy issues were involved. Therefore, the opinions of the ethics committee and patients were not considered.

2.2. Selection criteria

The systematic selection of eligible studies for this meta-analysis was based on the following conditions: retrospective or prospective, case-control or cohort studies, and clinical trials along with randomized controlled trials; superficial esophageal neoplasms were histologically proven; ESD (not EMR) was conducted; risk factors for postoperative stricture were analyzed; only English or Chinese language studies could be included. Ineligible studies included: stricture rates were not reported; studies had no control group; case reports, comments, reviews, letters, and surveys.

2.3. Data extraction

Two investigators extracted the data independent of the studies included in a data from that had been designed already. Discussion between 2 authors and consensus with a third one helped in conflict resolution of opinions. Information such as the first author's name, year of publication, country, study design, patient factors (number of patients their gender and age), the definition of stricture, lesion characteristics (location, macroscopic type, depth of invasion, longitudinal length, circumferential diameter, circumferential range, and histological type), procedural details (procedure time, perforation, delayed hematoma, and en bloc resection), and postoperative characteristics (postoperative bleeding, postoperative perforation, postoperative fever) were collected.

2.4. Quality assessment

For quality check, 2 individual investigators analyzed the included studies critically following the Newcastle-Ottawa Scale (NOS),^[27] which does the quality assessment of cohort and case-control studies. Any disagreements were resolved through discussion. Its selection criteria comprise of assessing comparability and exposure with a maximum score of 9. Studies with NOS scores ≥6 were regarded as studies of high quality, whereas those with NOS scores of <6 were considered as studies of low quality.

2.5. Statistical analysis

Odds ratios (OR) and 95% confidence intervals (CI) for every categorical risk factor and the mean or median for continuous data were registered (or calculated at a given opportunity). This meta-analysis included studies with data allowing the calculation of OR, which was performed by computing pooled OR for categorical variables and weighted mean differences (WMD) for continuous variables. Statistical nonuniformity was assessed by using the Cochrane Q statistic and the I² statistic. For the Q-statistic, a P-value of <.10 was taken as the representative of significant statistical heterogeneity. I² ranged between 0% and 100% indicated the total variation percentage across studies due to the nonuniformity. An I² percentage of >50% may indicate substantial heterogeneity. The summary estimate based on the random-effects model was presented, as this provided a more narrowed and specific approximation of exposure effect where there was a high probability of considerable variance between studies (DerSimonian and Laird model).^[28] A fixed-effect model was implemented otherwise.

For the evaluation of the dose–response association between circumferential range and risk of stricture, the general model of the least-squares was described by Greenland and Longnecker^[29] and Orsini et al^[30] was used to analyze the trend from the interrelated estimates for log OR through classes of circumferential range. A restricted cubic spline with 3 knots at 3 percentiles 10%, 50%, and 90% of the distribution was modeled.^[31] A P-value for curve linearity or nonlinearity of the curve was measured by testing the null hypothesis of the coefficient of the second spline being equal to 0.^[32] When the nonlinear trend was left undetected, linear analysis was done using the above-mentioned procedure.^[29] The median or mean circumferential range was assigned to every study for the category to individual corresponding OR. Upon getting no report output of median or mean range, the midpoint of the upper and lower boundaries in every individual category was assigned as the median range. If the upper boundary was missing for the highest category, the category midpoint was set at 1.5X the lower boundary. For the open-ended lowest category, the lower boundary was set to 0. In the event of nil outcomes, a correction factor of 0.5 for continuity correction.

Subgroup analyses were done for the detection of the cause of heterogeneity and evaluation of potential effect between the depth of tumor invasion and circumferential degree. Visual inspection was done for the evaluation of publication bias by a funnel plot, and Egger regression asymmetry test helped in the further judgment of the publication bias.^[33] A P-value of <.05 was taken as a demonstration of major statistical publication bias. Egger test and funnel plot are inadequate when the studies are <10.^[34] The software Stata version 15.1 (Stata Corp, College Station, TX) was used to conduct all meta-analyses. This study was registered with PROSPERO, number CRD42019128111.

3. Results

3.1. Identification of studies

The literature review and process of selecting eligible studies was shown in Fig. 1. At the start 611 articles related to our topic were discovered, 238 out of which were eliminated because they were reproduced from other studies. After close analysis of abstracts and titles, 305 articles were eliminated. Based on the assessment of complete text, a total of 11 studies^[35–45] were found eligible for inclusion in the meta-analysis.

Figure 1.

Flow diagram of literature search and study selection.

3.2. Study characteristics

Among 11 studies included, 5 of them^{[36,37,39,42,44]} were cohort studies, and others were case-control studies. Table 1 provides the main features of included studies in the analysis. These studies were published between 2009 and 2018. Three of them^[35,36,38] were from Japan, and 8 were from China. The sample size ranged from 42 to 401, with 2248 as a total size (284 of the included studies were structure cases and 1964 were considered as controls).

Table 1.

Characteristics of studies included in the meta-analysis.

Reference	Year	Country	Study design	Timeframe	Cases	Controls	Overall PSR (%)	NOS(score)
Mizuta et al[35]	2009	Japan	Case-control study	2003–2008	7	35	16.7	7
Ono et al[36]	2009	Japan	Cohort study	2002–2008	11	54	16.9	7
Ju et al[37]	2013	China	Cohort study	2007–2012	42	320	11.6	7
Funakawa et al[38]	2015	Japan	Case-control study	2009–2013	22	98	18.3	7
Zhuang et al[39]	2016	China	Cohort study	2009–2014	11	84	11.6	7
Cao et al[40]	2017	China	Case-control study	2014–2016	33	250	11.7	7
Dong et al[41]	2017	China	Case-control study	2014–2016	26	171	13.2	9
Fan et al[42]	2018	China	Cohort study	2013–2015	36	265	12.0	7
Yang et al[43]	2018	China	Case-control study	2011–2017	22	110	16.7	7
Liu et al[44]	2018	China	Cohort study	2011–2016	53	331	13.8	7
Lu et al[45]	2018	China	Case-control study	2009–2016	21	146	12.6	9

NOS = Newcastle-Ottawa Scale, PSR = postoperative stricture rate.

3.3. Quality assessment

The risk of bias was assessed in the included studies making use of the NOS. Cases in included studies were chosen carefully, as only surgically confirmed cases were recruited (Table 2), and disease extent was explained in detail. None of the selected studies had an average score of 7.4 and none had a score <7.

Table 2.

Quality scores of included studies.

	Selection				Comparability	Exposure
Reference	1	2	3	4	5	6	7	8	Score
Mizuta et al[35]	1	1	1	1	0	1	1	1	7
Ono et al[36]	1	1	1	1	0	1	1	1	7
Ju et al[37]	1	1	1	1	0	1	1	1	7
Funakawa et al[38]	1	1	1	1	0	1	1	1	7
Zhuang et al[39]	1	1	1	1	0	1	1	1	7
Cao et al[40]	1	1	1	1	0	1	1	1	7
Dong et al[41]	1	1	1	1	2	1	1	1	9
Fan et al[42]	1	1	1	1	0	1	1	1	7
Yang et al[43]	1	1	1	1	0	1	1	1	7
Liu et al[44]	1	1	1	1	0	1	1	1	7
Lu et al[45]	1	1	1	1	2	1	1	1	9

3.4. Meta-analysis

Risk factors for postoperative stricture after ESD were explored by all the included studies. Identified risk factors were clustered into 4 themes (patient factors, lesion characteristics, procedural details, and postoperative characteristics). All potential risk factors were shown in Table 3, and major risk factors from the meta-analysis were highlighted in bold. Risk factors of categorical variables were summarized in Fig. 2, while those of continuous variables were summarized in Fig. 3.

Table 3.

Meta-analysis of potential risk factors among included studies.

Potential risk factors	No. of studies	Total sample	Statistic method	Pooled OR, WMD	I2 (%)
Patient factors
Age, y	9	1181	I–V, random	1.61 (−0.32–3.53)	73.3
Gender	10	2106	I–V, fixed	1.03 (0.77–1.36)	0
Drinking history	2	252	I–V, fixed	1.69 (0.79–3.62)	0
Hypertension	3	946	I–V, random	1.08 (0.47–2.48)	60.3
Diabetes	4	1078	I–V, fixed	1.37 (0.81–2.32)	0
Family history of esophageal cancer	3	895	I–V, fixed	1.39 (0.73–2.61)	0
Lesion characteristics
Tumor location
Upper vs middle/lower	11	2148	I–V, fixed	1.51 (1.02–2.25)	31.2
Middle vs Upper/lower	11	2148	I–V, fixed	0.99 (0.77–1.29)	0
Lower vs Upper/middle	11	2148	I–V, fixed	0.85 (0.65–1.13)	0
Macroscopic type
Elevated vs flat/depressed	3	294	I–V, fixed	1.71 (0.77–3.79)	0
Flat vs elevated/depressed	3	294	I–V, fixed	0.88 (0.38–2.03)	0
Depressed vs elevated/flat	3	294	I–V, fixed	0.62 (0.18–2.22)	0
Macroscopic type
IIb/IIc vs IIa	3	551	I–V, random	0.64 (0.16–2.58)	59.1
IIa/IIc vs IIb	3	551	I–V, fixed	2.76 (1.55–4.92)	0
IIa/IIb vs IIc	3	551	I–V, fixed	0.58 (0.27–1.25)	5.9
Tumor depth of invasion
m2/m3/sm1/sm2 vs m1	7	1677	I–V, random	7.47 (3.31–16.86)	76.3
m3/sm1/sm2 vs m1/m2	7	1677	I–V, random	12.67 (4.00–40.10)	88.9
Longitudinal length, mm	9	1733	I–V, random	13.75 (7.76–19.74)	84.4
Circumferential diameter, mm	3	624	I–V, fixed	10.87 (8.13–13.60)	0
Circumferential range
>3/4 vs <3/4	7	1423	I–V, random	38.17 (9.94–146.52)	86.5
>3/4 vs <1/2	5	1106	I–V, fixed	222.09 (72.75–677.97)	38.1
1/2–3/4 vs <1/2	5	1106	I–V, fixed	6.34 (2.83–14.20)	0
Histological type
HGIN vs carcinoma	4	605	I–V, fixed	0.72 (0.41–1.24)	18.8
Procedural details
Operating time, min	5	1227	I–V, random	26.23 (-1.52–53.98)	96.1
En bloc resection	3	687	I–V, fixed	0.92 (0.51–1.68)	0
Postoperative characteristics
Postoperative bleeding	3	758	I–V, fixed	1.29 (0.22–7.46)	0
Postoperative perforation	4	800	I–V, fixed	2.34 (0.81–6.76)	0
Postoperative fever	2	252	I–V, random	0.87 (0.18–4.27)	62.0

HGIN = high grade intraepithelial neoplasia.

Figure 2.

Risk factors of categorical variables for postoperative stricture after ESD. ESD = endoscopic submucosal dissection.

Figure 3.

Risk factors of continuous variables for postoperative stricture after ESD. ESD = endoscopic submucosal dissection.

3.5. The definition and rate of postoperative stricture

The most common definition of postoperative stricture was “a standard endoscope of 9.8 mm in diameter could not go through the stricture.” Some studies also classified the degree of stricture according to the diameter of the esophagus. Postoperative stricture rates ranged from 9.0%^[42] to 18.3%^[38] and the rate of pooled postoperative stricture rate was 12.2% (95% CI: 10.9–13.6%) in the fixed effects model, with low significant heterogeneity across studies (I² = 22.5%, P = .229). Postoperative stricture rates did not show country-wise variability (17.6% in studies of Japan and 11.8% in studies of China, I² = 22.5%).

3.6. Patient factors

Nine studies considered patient age could be a potential risk factor.^{[36–39,41–45]} The meta-analysis showed the WMD of 1.61 years (95% CI: –0.32–3.53) in the random-effects model, indicating that age had no role in increasing the risk of postoperative stricture, although the study finding had substantial heterogeneity (I² = 73.3%).

Gender was estimated as a potential risk factor by 10 studies.^[36–45] The meta-analysis showed that gender did not play any role in the enhanced risk of postoperative stricture (OR 1.03, 95% CI: 0.77–1.36, I² = 0%).

Two studies evaluated drinking history as a potential risk factor for postoperative stricture.^[38,43] The meta-analysis indicated that was not the case and drinking history did not increase the risk of postoperative stricture (OR 1.69, 95% CI: 0.79–3.62, I² = 0%).

Three studies suggested that hypertension may be a potential risk factor.^[37,40,43] The meta-analysis gave a contrast result of hypertension not being associated with an increased risk of postoperative stricture (OR 1.08, 95% CI: 0.47–2.48, I² = 60.3%).

Four studies evaluated diabetes as a potential risk factor.^{[37,40,42,43]} Diabetes was not associated with an increased risk of postoperative stricture according to the meta-analysis (OR 1.37, 95% CI: 0.81–2.32, I² = 0%).

Three studies evaluated a family history of esophageal cancer as a potential risk factor.^[37,42,43] Opposed to that, this meta-analysis found that family history did not have a role in an increased risk of postoperative stricture (OR 1.39, 95% CI: 0.73–2.61, I² = 0%).

3.7. Lesion characteristics

All studies evaluated tumor location as a potential risk factor. The meta-analysis indicated that tumors present in the upper third of the esophagus had substantial differences as compared with the middle/lower third (OR 1.51, 95% CI: 1.02–2.25, I² = 31.2%). In addition, tumors located in the middle third of the esophagus were not found to be an increased risk of postoperative stricture (OR 0.99, 95% CI: 0.77–1.29, I² = 0%). There was also no major variation in postoperative stricture between tumors in the lower third of the esophagus and the upper/middle third (OR 0.85, 95% CI: 0.65–1.13, I² = 0%).

Five studies^{[35,36,38,43,44]} evaluated macroscopic type as a potential risk factor. In 3^[35,38,43] of these studies, the types of elevated, flat, and depressed were analyzed. Three^[36,38,44] studies also provided the data of 3 subtypes of flat type. The meta-analysis indicated no substantial variations in postoperative stricture among elevated type and flat/depressed type (OR 1.71, 95% CI: 0.77–3.79, I² = 0%), and also between flat type and elevated/depressed type (OR 0.88, 95% CI: 0.38–2.03, I² = 0%), as well as between depressed type and elevated/flat type (OR 0.62, 95% CI: 0.18–2.22, I² = 0%). The flat type was further divided into 3 subtypes: flat with elevation (IIa), flat without elevation or depression (IIb), and flat with depression (IIc). The meta-analysis indicated that macroscopic type of IIa/IIc was more likely to form postoperative stricture than IIb (OR 2.76, 95% CI: 1.55–4.92, I² = 0%). And there were no statistical significances in IIb/IIc versus IIa (OR 0.64, 95% CI: 0.16–2.58, I² = 59.1%) and IIa/IIb versus IIc (OR 0.58, 95% CI: 0.27–1.25, I² = 5.9%).

Tumor depth of invasion was evaluated as a possible risk factor in 7 studies.^{[36–40,43,44]} The meta-analysis indicated that invasion depth was substantially linked with postoperative stricture, comparing m₂/m₃/sm₁/sm₂ versus m₁ (OR 7.47, 95% CI: 3.31–16.86, I² = 76.3%) and m₃/sm₁/sm₂ versus m₁/m₂ (OR 12.67, 95% CI: 4.00–40.10, I² = 88.9%). When studies were sub-grouped based on the different circumferential range for ESD (<3/4 vs >3/4),^[36,37] tumor depth to m₂ and more was still significantly associated with postoperative stricture in both subgroups (OR 19.01, 95% CI: 6.10–59.30, I² = 0%; OR 9.61, 95% CI: 1.93–47.94, I² = 0%) (Table 4). The results showed that a high tumor depth of invasion was associated with a greater chance of postoperative stricture.

Table 4.

Subgroup analysis of tumor depth of invasion and circumferential range.

Variable	Stratification criterion	No. of studies	Pooled OR	I2 (%)	P value
Tumor depth of invasion	Overall	2	15.61 (6.20,39.29)	0	.000
Circumferential range	<3/4	2	19.01 (6.10,59,30)	0	.000
	>3/4	2	9.61 (1.93,47.94)	0	.006
Circumferential range	Overall	2	15.90 (6.57,38.50)	0	.000
Tumor depth of invasion	m1	2	34.58 (7.42,161.25)	1.8	.000
	m2/m3/sm1/sm2	2	13.71 (5.04,37.24)	0	.000

Longitudinal length was evaluated as a possible risk factor in 9 studies.^{[35–39,41,42,44,45]} The meta-analysis found a WMD of 13.75 mm (95% CI: 7.76–19.74) in the random-effects model, indicating that longer longitudinal diameter was related to an enhanced risk of postoperative stricture, although there was considerable heterogeneity between study findings (I² = 84.4%).

The circumferential diameter was evaluated as a possible risk factor in 3 studies.^[36,37,41] The meta-analysis found a WMD of 10.87 mm (95% CI: 8.13–13.60, I² = 0%) in the fixed effects model, showing that longer circumferential diameter was associated with an increased risk of postoperative stricture.

The circumferential range was evaluated as a possible risk factor in 7 studies.^[36–42] The meta-analysis showed major variation between circumferential range >3/4 versus <3/4 (OR 38.17, 95% CI: 9.94–146.52, I² = 86.5%), >3/4 versus <1/2 (OR 222.09, 95% CI: 72.75–677.97, I² = 38.1%) and 1/2 to 3/4 versus <1/2 (OR 6.34, 95% CI: 2.83–14.20, I² = 0%). When studies were subgrouped based on different tumor depth of invasion (m₁ vs m₂/m₃/sm₁/sm₂),^[36,37] three-fourth or less of the circumference of esophagus was still significantly related to postoperative stricture in both subgroups (OR 34.58, 95% CI: 7.42–161.25, I² = 1.8%; OR 15.90, 95% CI: 6.57–38.50, I² = 0%) (Table 4). Six studies were a part of analyzing the dose–response of circumferential range and stricture risk.^{[36,37,39,40,42,44]} Upon using the restricted cubic splines model, no evidence of a curve linear association between the 2 was observed (P = .92 for nonlinearity). Results demonstrated that each additional 10% of the circumferential range increased the risk of stricture by 149% (OR = 9282.46, 95% CI: 978.14–88089.35, P = .000) (Fig. 4).

Figure 4.

Dose–response association between circumferential range and risk of postoperative stricture after ESD. The solid line and long dashed lines represent the estimated OR and it is 95% CI. CI = confidence intervals, ESD = endoscopic submucosal dissection, OR = odds ratios.

3.8. Procedural details

Five studies were evaluated for the operating time being a potential risk factor.^{[36–38,42,44]} The meta-analysis used random-effects models and suggested a WMD of 26.23 minutes (95% CI: –1.52–53.98), indicating no association of operating time with an increased risk of postoperative stricture, even though the study results had substantial heterogeneity (I² = 96.1%).

Three studies evaluated en bloc resection as a potential risk factor.^[35,36,40] The meta-analysis showed no connection of en bloc resection with the enhanced risk of postoperative stricture (OR 0.92, 95% CI: 0.51–1.68, I² = 0%).

3.9. Postoperative characteristics

Three studies evaluated postoperative bleeding as a potential risk factor.^[37,39,42] The meta-analysis did no support this finding and suggested no relatedness of postoperative bleeding to an elevated risk of postoperative stricture (OR 1.29, 95% CI: 0.22–7.46, I² = 0%).

Four studies evaluated postoperative perforation as a potential risk factor.^{[35,37,39,42]} The meta-analysis opposed these findings by showing no connection of postoperative perforation with a high risk of postoperative stricture (OR 2.34, 95% CI: 0.81–6.76, I² = 0%).

Two studies evaluated postoperative fever as a potential risk factor.^[38,43] The meta-analysis showed different findings that postoperative fever was not related to the high risk of postoperative stricture (OR 0.87, 95% CI: 0.18–4.27, I² = 62.0%).

3.10. Publication bias

The meta-analysis involved >10 studies. Estimation of potential publication bias was done by examining funnel plots qualitatively and Egger tests quantitatively. The funnel plot and Egger test (P = .079 and P = .931, respectively) indicated no proof of publication bias in the comparison of age and tumor location for postoperative stricture (Figs. 5 and 6).

Figure 5.

Egger funnel plot for publication bias analysis for age and risk of postoperative stricture after ESD. ESD = endoscopic submucosal dissection.

Figure 6.

Egger funnel plot for publication bias analysis for tumor location and the risk of postoperative stricture after ESD. ESD = endoscopic submucosal dissection.

4. Discussion

This meta-analysis evaluated risk factors associated with postoperative stricture after endoscopic submucosal dissection for superficial esophageal neoplasms. It is a noteworthy issue because some studies detect different risk factors for postoperative stricture. All studies comprising this meta-analysis reported postoperative stricture after ESD to occur in 9.0% to 18.3%, and the pooled rate was 12.2% after our review. It showed that postoperative stricture was a common complication which we should focus on. NOS was used in this meta-analysis for checking the quality of case-control. The highest score of NOS is 9, and scores <6 indicated a low-quality study. As shown in Table 2, the quality assessment had scores up to 6 and not less. We are certain that the clinical evidence would be better presented by this quality assessment and help in better assessment of the results’ reliability.

The meta-analysis demonstrated that 6 risk factors of lesion characteristics showed substantial relatedness with postoperative stricture, including tumor location, macroscopic type, tumor depth of invasion, longitudinal length, circumferential diameter, and circumferential range. While other risk factors of patient factors, lesion characteristics, procedural details, and postoperative characteristics were not considered as prominent risk factors for postoperative stricture.

The meta-analysis found that a tumor located in the upper third of the esophagus was linked with an increased risk of postoperative stricture. Nevertheless, the majority of the included studies found that tumor location was not related to postoperative stricture, while 2 studies^[36,42] reported that postoperative stricture more likely occurred in the upper third of the esophagus than in the lower esophagus. The finding supports the fact that the luminal diameter of the esophagus is variable among different levels. The luminal diameter of the upper esophagus is smaller as compared with the lower esophagus. So it is can be inferred that the smaller diameter may develop an enhanced risk of postoperative stricture after ESD.^[46] However, extensive research should be carried out to verify the strength of the connection between tumor location and the occurrence of postoperative stricture because the sample of tumor located in the upper third of the esophagus was small. Also, the meta-analysis indicated that the macroscopic type of IIa/IIc was more likely to form postoperative stricture than IIb, which should be handled attentively because of the fewer cases.

After the meta-analysis, tumor depth to m₂ and deeper was thought to be a significant risk factor for postoperative stricture in total or subgroup analyses. When the depth of tumor infiltration exceeds m₃ or sm₁, refractory stricture more likely occurred after ESD.^[47] Mizuta et al^[35] and Funakawa et al^[38] noted that invasion depth was not related to the stricture, while other findings suggested it to be an independent risk factor. We speculate that the variability could be credited to diverse patient characteristics in the studies, which might be affected by other risk factors. Better research is required for the confirmation of the influence of tumor depth on postoperative stricture.

In the meta-analysis, it was found that longitudinal length was longer in the stricture group with a statistically significant difference. Several studies had confirmed that whatever the longitudinal length of the lesion was >30, 40, or 50 mm, it had an obvious correlation with postoperative stricture after ESD.^{[24,36,40,43,48,49]} Ono et al^[4] and Tang et al^[47] revealed that lesions of longitudinal mucosal defect length >50 mm after entire circumferential esophageal ESD were significant risk factors for refractory stricture.

Circumferential diameter and range were considered to be closely associated with post-ESD stricture in the study. It was found that the risk of postoperative stricture increased with the increase of circumferential diameter and range, especially when the circumferential mucosal defect was more than three-fourths, and these findings were backed up by this meta-analysis. It was first proposed by Mizuta et al^[35] that the predictive factor of post-ESD esophageal stricture as being a circumferential mucosal defect size, with a cut-off value of 71% yielding the highest sensitivity (100%) and specificity (97.1%). Most of the subsequent studies adopted 1/2 and 3/4 of the esophageal circumference as a classification to assess risk factors, which was also convenient for clinical application and demonstrated that resecting areas larger than 3/4 of the esophageal circumference was considered as an important independent risk factor for postoperative stricture. In the meta-analysis, comparing with <1/2 of the esophageal circumference, the increased risk of stricture was statistically significant for 1/2 to 3/4 of the esophageal circumference. However, considering that the overall incidence of stricture occurred in lesions with the circumferential extension of 1/2 to 3/4 was approximately 4% to 28%, it was suggested that excessive intervention was not necessary, and a follow-up observation was generally recommended after ESD.^{[4,36,37,42,50]} When the area of periesophageal mucosal defect exceeding 3/4 of the esophageal circumference, the incidence of stricture could be as high as 40% to 95%. In particular, postoperative stricture almost always occurred in patients who underwent entire circumference ESD.^[4,36] As we best understand, this is the first meta-analysis study on dose–response that evaluated the relationship between circumferential range and the occurrence of postoperative stricture after ESD. Based on our meta-analysis, which included 6 observational studies involving 186 cases and 1490 participants, our data suggest that a significant linear dose–response relationship was observed, and the risk of post-ESD stricture was significantly increased by 149% for every 10% increment of the circumferential range.

There are several potential mechanisms to explain postoperative stricture formation after ESD. The traditional theories consider esophageal stricture as a consequence of esophageal wall damage followed by scar tissue development and contraction of the secondary tissue.^[51] Some researchers suggested that withering of the muscularis propria and fibrosis of the submucosa were key factors in postoperative stricture after ESD.^[49,52] Honda et al^[53] reported that withering of the muscularis propria during the wound healing process was associated with stricture after esophageal EMR in a dog model. Kakushima et al^[54] observed fibrosis and thickening of the wall after 2 weeks of ESD, with complete stricture requiring almost 2 weeks. Miwata et al^[55] showed that damage to muscle layer during ESD was a key factor related to refractory stenosis group, and muscle layer damage may exacerbate atrophy of the muscularis propria and esophageal stricture. Lesions with large diameter, deep infiltration, and thermal damage of deep tissue during operation would lead to the formation of a stricture in the above ways. Meanwhile, Dong et al^[41] found that patients with excessive longitudinal diameter and circumferential diameter were more difficult to operate on and had more intraoperative bleeding. If satisfactory hemostasis was not achieved promptly, it would lead to vascular obstruction and postoperative stricture. Zhang et al^[56] reported that the cautery time was longer, and the local thermal damage was more serious after active bleeding, which was considered to be possibly correlated with post-ESD stricture. Liu et al^[51] suggested that direct relatedness of esophageal stricture formation to the loss of the submucosa and instead of damage to it and shared the view that mucosal loss was a key mechanism for the occurrence of esophageal stricture following ESD in the new study. In addition, it was mentioned that the occurrence of stricture may be different in different study populations influenced by host factors, such as environmental factors, single-nucleotide polymorphisms, and other habits such as dietary routine, etc.^[38] In the light of the possible mechanisms mentioned above, we highly suggested the following to prevent post-ESD stricture based on complete resection of the lesion. Firstly, dissection should be attentively performed while holding the muscle layer structure because of the damage to the muscularis propria during submucosal dissection in most cases. Secondly, maximum possible mucosa should be retained during ESD procedures. Thirdly, in the case of large-scale esophageal ESD, longitudinal diameter and circumferential range of mucosal dissection should be minimized.

This study was deemed to be of several strengths. Firstly, this was the first meta-analysis focusing on risk factors of postoperative stricture formation in patients who received ESD and took a wide range of indicators, regarding patient factors, lesion characteristics, procedural details, and postoperative characteristics to comprehensively assess possible risk factors. Secondly, this was the first meta-analysis regarding this subject to explore the dose–response relationship between circumferential range and post-ESD stricture, which provided better quantification accuracy. Thirdly, good quality articles were chosen for this meta-analysis and they reported elaborate baseline characteristics of participants. Fourthly, subgroup analysis gave uniform results and significant publication bias was absent. This supported our findings to be robust and reliable.

However, we must acknowledge the limitations of the meta-analysis. Firstly, the analysis was completely based on observational studies of high quality but it might introduce potential selection bias and recall bias. Secondly, the studies included in this analysis were from Japan and China, and the search of the literature was limited in language to English and Chinese, which might exclude publications in other languages. Thirdly, the small number of cases included in the study can also be a bias, but this occurs because of the limited number of included studies. Therefore, our results need further confirmation in other countries. Fourthly, we failed in the identification of interactions and confound between risk factors, although we believed that given a large amount of included patients and the low heterogeneity in most of the evaluated risk factors, it allowed us to draw some important conclusions. Fifthly, due to the small number of included studies, the heterogeneity that occurred when some indicators were pooled was not fully explained.

5. Conclusion

In summary, the meta-analysis pointed out 6 risk factors related to the enhanced risk of postoperative stricture after ESD for superficial esophageal neoplasms, containing the upper third of the esophagus, macroscopic type of IIa/IIc, tumor depth of invasion above m1 and m2, longitudinal length, circumferential diameter, circumferential range >3/4. The relationship was linear for circumferential range and post-ESD stricture. The results can help doctors identify patients at increased risk and thus direct management in terms of the adequate period of surveillance time post-ESD and take available approaches of stricture prevention. To learn more about post-ESD stricture for superficial esophageal neoplasms, more detailed research is suggested in the future, including a detailed analysis of subgroups and additional data on the conditions of the occurrence of strictures.

Acknowledgments

The authors acknowledge the databases (The PubMed, MEDLINE, EMBASE, Web of Science, Cochrane Central Register of Controlled Trials, as well as China National Knowledge Infrastructure, the Wanfang Database, and the Chinese Biomedical Literature Database) for providing their platforms and contributors for uploading their meaningful datasets.

Author contributions

Conceptualization: Nan Lin, Jie Lin, Jinrong Gong.

Data curation: Nan Lin, Jie Lin, Jinrong Gong.

Formal analysis: Nan Lin, Jie Lin.

Investigation: Nan Lin, Jie Lin.

Methodology: Nan Lin, Jie Lin, Jinrong Gong.

Resources: Nan Lin, Jie Lin.

Software: Nan Lin, Jie Lin.

Supervision: Jinrong Gong.

Writing – original draft: Nan Lin, Jie Lin.

Writing – review & editing: Nan Lin, Jinrong Gong.