CO2 insufflation versus air insufflation for endoscopic submucosal dissection: A meta-analysis of randomized controlled trials

Background Carbon dioxide (CO2) insufflation is increasingly used for endoscopic submucosal dissection (ESD) owing to the faster absorption of CO2 as compared to that of air. Studies comparing CO2 insufflation and air insufflation have reported conflicting results. Objectives This meta-analysis is aimed to assess the efficacy and safety of use of CO2 insufflation for ESD. Methods Clinical trials of CO2 insufflation versus air insufflation for ESD were searched in PubMed, Embase, the Cochrane Library and Chinese Biomedical Literature Database. We performed a meta-analysis of all randomized controlled trials (RCTs). Results Eleven studies which compared the use of CO2 insufflation and air insufflation, with a combined study population of 1026 patients, were included in the meta-analysis (n = 506 for CO2 insufflation; n = 522 for air insufflation). Abdominal pain and VAS scores at 6h and 24h post-procedure in the CO2 insufflation group were significantly lower than those in the air insufflation group, but not at 1h and 3h after ESD. The percentage of patients who experienced pain 1h and 24h post-procedure was obviously decreased. Use of CO2 insufflation was associated with lower VAS scores for abdominal distention at 1h after ESD, but not at 24h after ESD. However, no significant differences were observed with respect to postoperative transcutaneous partial pressure carbon dioxide (PtcCO2), arterial blood carbon dioxide partial pressure (PaCO2), oxygen saturation (SpO2%), abdominal circumference, hospital stay, white blood cell (WBC) counts, C-Reactive protein (CRP) level, dosage of sedatives used, incidence of dysphagia and other complications. Conclusion Use of CO2 insufflation for ESD was safe and effective with regard to abdominal discomfort, procedure time, and the residual gas volume. However, there appeared no significant differences with respect to other parameters namely, PtcCO2, PaCO2, SpO2%, abdominal circumference, hospital stay, sedation dosage, complications, WBC, CRP, and dysphagia.


Results
Eleven studies which compared the use of CO 2 insufflation and air insufflation, with a combined study population of 1026 patients, were included in the meta-analysis (n = 506 for CO 2 insufflation; n = 522 for air insufflation). Abdominal pain and VAS scores at 6h and 24h post-procedure in the CO 2 insufflation group were significantly lower than those in the air insufflation group, but not at 1h and 3h after ESD. The percentage of patients who experienced pain 1h and 24h post-procedure was obviously decreased. Use of CO 2 insufflation was associated with lower VAS scores for abdominal distention at 1h after ESD, but not at 24h after ESD. However, no significant differences were observed with respect to postoperative transcutaneous partial pressure carbon dioxide (PtcCO 2 ), arterial blood carbon dioxide partial pressure (PaCO 2 ), oxygen saturation (SpO 2 %), abdominal circumference, hospital PLOS

Introduction
Endoscopic submucosal dissection (ESD) allows lesions to be dissected and resected directly along the submucosal (sm) layer with use of an electrosurgical knife. With rapid advances in endoscopic techniques, ESD has become an invaluable tool in the treatment of early neoplasms of the gastrointestinal tract, particularly for large lesions [1]. However, this procedure is timeconsuming and requires special endoscopic experience. The incidence of complications such as perforations, hemorrhage, and abdominal discomfort has also increased as a direct result [2]. Insufflation is required to achieve adequate visualization during ESD. Postoperatively, the gas is not absorbed immediately and remains in the gastrointestinal tract, which can cause abdominal pain and distension. Carbon dioxide (CO 2 ) is rapidly cleared from the small bowel and excreted through the lungs, thus allowing the bowel to deflate quickly [3]. The benefits of use of CO2 over air were first pointed out by Rogers [4]. It is generally believed that use of CO 2 insufflation is associated with less severe pneumoperitoneum, abdominal pain and abdominal distension [5,6]. Recent clinical studies have shown that CO 2 insufflation for ESD is safe and effective; however, Maeda et al [7] observed no significant difference with respect to post-procedural abdominal pain or discomfort between the CO 2 insufflation and air insufflation groups. In addition, two meta-analyses of studies which compared the use of CO 2 insufflation versus air insufflation for gastrointestinal endoscopy and endoscopic retrograde cholangiopancreatography (ERCP) reinforced the advantages of CO 2 insufflation (lower post-procedural pain and bowel distension); however, they did not find any advantage with respect to arterial blood CO 2 partial pressure (PaCO 2 ) and transcutaneous partial pressure CO 2 (PtcCO 2 ) levels [8,9].
Since then, numerous randomized controlled trials (RCTs) have addressed the role of abdominal discomfort and CO 2 variation; however, other advantages of use of CO 2 insufflation have not been adequately investigated. Few RCTs have assessed comprehensive indicators, such as the dosage of sedative drugs, procedure time and incidence of complications. Additionally, the number of patients in these RCTs were largely inadequate. Although the superiority of CO 2 insufflation with respect to postoperative abdominal discomfort is backed by strong evidence from RCTs, the safety and efficacy of CO 2 insufflation for ESD treatment has not been assessed by a meta-analysis. We, therefore, sought to assess the safety and efficacy of use of CO 2 insufflation in patients undergoing ESD by performing a meta-analysis.
Chinese Biomedical Literature Database (Sinomed). All studies published as of January 2017 were eligible for inclusion. No restriction was imposed with respect to the language of publication or type of article. The following free-text terms and MeSH terms were used to retrieve studies: "carbon dioxide," "ESD," and "air". The search strategy is summarized in S1 Table. Study selection and data extraction Only RCTs that compared the use of CO 2 insufflation with that of insufflation for ESD were selected. Two reviewers (Li X & Dong H) independently summarized information and data from each study using a standardized format. Any disagreement over study selection was resolved by referring to the adjudicating senior author (Zhang GX). Duplicate articles were excluded with the use of the software package Endnote X4 (reference management software). The title and abstract of the selected articles were screened to exclude articles that did not qualify the inclusion criteria, followed by a full text review of all eligible articles. In the event of lack of original data in the article, the respective authors were contacted to request further information.

Outcomes
The primary outcomes for this analysis were (i) visual analogue scale (VAS) score for pain; (ii) percentage of patients without pain at various time points after ESD; and (iii) PtcCO 2 and PaCO 2 levels at different time points after ESD. The secondary outcomes were: abdominal distention, abdominal circumference, oxygen saturation (SpO 2 %), total procedure time; average hospital stay; dose of sedative drugs; incidence of complications (pneumonia, hemorrhage, perforation, and emphysema); clinical course (anal exsufflation, dysphagia, residual gas, and Mallory-Weiss tear); laboratory examination (white blood cell count [WBC] and C Reactive Protein [CRP] level).

Quality assessment
Two authors (Li X and Dong H) independently evaluated the methodological quality of the included studies with use of the 'risk of bias' assessment tool from the Cochrane Handbook for Systematic Reviews of Interventions [10]. The following aspects of the methodology were assessed: random sequence generation, concealment of allocation, blinding of subjects and personnel, blinding of outcome assessment, completeness of reporting of outcomes data, potential reporting bias, and other sources of bias. The studies were divided into three groups based on the assessed risk of bias: high risk of bias, low risk of bias, and unclear. Disagreements were resolved by referring to the third author (Zhang GX).

Statistical analysis
Meta-analyses were conducted with Review Manager 5.3 (Cochrane Collaboration, Oxford, UK). The random-effects model was used for all analyses owing to clinical heterogeneity among the selected studies [11]. The outcome variables were dichotomous. Risk ratios (RR) with 95% confidence intervals (CIs) were calculated by Mantel-Haenszel method. For continuous variables, mean difference (MD) with 95% CIs were calculated with inverse variance method [12]. For continuous outcomes measured on different scales, standardized mean difference (SMD) with 95% CI is reported. Heterogeneity among the selected trials was assessed with I 2 measure of inconsistency (cutoff level of I 2 = 50%). Sensitivity analyses were performed only for high-quality studies. If sufficient data were available, publication bias was assessed with use of funnel plots.

Identification and characteristics of studies
A total of 37 records were retrieved on initial search on electronic databases, including The Cochrane Library (n = 7), PubMed (n = 9), EMBASE (n = 12), Chinese Biomedical Literature Database (n = 6), and manual search of the references of the included RCTs (n = 1). Eighteen duplicates and four non-relevant articles were excluded after review of titles and abstracts. After a review of full texts, 2 articles with retrospective study design and an uncontrolled study were excluded. Finally, eleven publications were included in this meta-analysis; of these, ten were original research articles published in journals while one was a conference abstract. The characteristics of included studies are summarized in Table 1. A total of 1026 patients were involved; 506 were administered CO 2 insufflation and 522 were administered air insufflation. In the third trial, 30 participants (12 from the CO 2 group and 18 from the air group) particularly underwent further analysis, including PaCO 2 , procedure time, and sedative drugs dosage. All studies were conducted in Asia: eight in Japan [7,[13][14][15][16][17][18], three in China [19][20][21], and one in South Korea [22]. ESD equipment used in the trials included Olympus Optical Co [13,15], Olympus GIF-Q260J [18,21,22], Olympus XGIF-2T240 M [7], Olympus 165 [14], Pentex 2940 [19], and Olympus CV-260SL [20]. For two articles, relevant details were not mentioned [16,17].

Primary outcomes: Abdominal pain.
Pain VAS score and percentage of patients without pain was recorded at the following time-points: 1 h, 3 h, 6 h, and 24 h after ESD. Subgroup analyses was performed to assess these two outcomes. (i)Pain VAS score: Six articles reported pain VAS scores at 1 h and 24 h after ESD [7,16,[18][19][20]22]; three articles reported VAS scores at 3h and 6h after ESD [7,16,18]. The meta-analysis revealed that pain VAS scores at 6h and 24h post-procedure in the CO 2 insufflation group were significantly lower than those in the air insufflation group (P = 0.0003 and P<0.00001, respectively). The difference in VAS scores at 1h and 3h after ESD were not statistically significant ( Fig 3A). (ii)Percentage of patients without pain: Maeda et al [7]and Kim et al [22] reported the post-procedural pain experience as percentage of patients who did not experience pain at 1h, 3h, and 24h after ESD. Feng et al [21]only reported the rate of pain-free patients at 1h post-procedure. No significant higher proportion of patients was observed in the air insufflation group at 3h after ESD; however, the number of patients with pain was significantly smaller in CO 2 insufflation group at 1h and 24h after ESD (Fig 3B).
2.6 Secondary outcomes: Laboratory examination. No significant between-group difference was observed with respect to WBC counts and CRP levels on day-1 after procedure in two studies [14] (Fig 4J and 4K).  [7, 13-15, 18, 20, 22], which included propofol, midazolam, and morphia. Only Saito et al [15] used midazolam to maintain the depth of sedation, while Takada et al [14] used midazolam in combination with diazepam and pentazocine; the others [13,18,20,22]were all combined other drugs to general anesthesia. Compared to air insufflation, dosage of sedative drugs did not obviously decrease in CO 2 insufflation group, except when midazolam was used alone (Fig 4L-4N).
2.8 Secondary outcomes: Clinical course. The included studies reported the following clinical parameters: residual gas, dysphagia, and Mallory-Weiss tear. Maeda et al reported dysphagia and gas volume in the gastrointestinal tract [7,18]. The residual gas volume in the CO 2 insufflation group was significantly lower than that in the air insufflations group (P< 0.00001, Fig 4O); dysphagia occurred in three patients in the air group and two patients in the CO 2 group (Fig 4P). Mallory-Weiss tear was described by Takada et al [14] and Onogi et al [17]; the incidence of Mallory-Weiss tear in the CO 2 group was significantly lower than that in the air group (0% vs. 15.6%, P = 0.013) [14]. Onogi et al [17] also reported lower incidence with use of CO 2 insufflation; however, specific data has not been reported.

Subgroup analysis and sensitivity analysis
As for heterogeneity with abdominal pain, abdominal distention and dosage of midazolam, we performed analysis of subgroup on the basis of a combination of drugs, country and type of Endoscope. However, the heterogeneity could not totally eliminate. It may be due to the sedation degree, the skill of endoscopist, etc. We could not conduct subgroup analyses on them as lack of original data. Only two studies [13,14] reported abdominal distention, which hardly to accomplish the subgroup analysis.
On sensitivity analysis with regard to abdominal pain VAS scores at 1h and 24h post-procedure, heterogeneity decreased slightly after exclusion of the data from the study by L Zhan et al [20] (Tables 2 and 3). It can be speculated that an assistant who was not blinded to the gas being used may have been involved in the outcomes assessment.

Discussion
In this meta-analysis, we assessed the safety and efficacy of CO 2 insufflation versus that of air insufflation for ESD treatment. The key finding of our analysis is that use of CO 2 insufflation was associated with lower abdominal pain and distension in the postoperative period. Moreover, use of CO 2 insufflation was associated with shorter procedure time. Finally, use of CO 2 insufflation offered a distinct advantage of lower residual gas volume in the gastrointestinal tract. However, no advantage of CO 2 insufflation was observed with respect to CO 2 retention, SpO 2 %, the length of hospital stay and abdominal circumference. Overall, this meta-analysis suggests that CO 2 insufflation is a safe and effective alternative to use of air insufflation for ESD. We mainly assessed the extent of post-procedural abdominal discomfort and CO 2 retention. Gas is deliberately insufflated into the gut lumen during ESD to facilitate visualization. In comparison with air, the gastrointestinal mucosa absorbs CO 2 faster, which is subsequently eliminated via the lung. Therefore, CO 2 insufflation possibly decreases the duration of abdominal distension [25]. However, Maeda et al [7] revealed that CO 2 insufflation was irrelevant to the subjective pain and distension of patients. An RCT [14] showed no effect on the number of patients without pain. The results of our meta-analysis suggest that CO 2 insufflation alleviated abdominal pain for at least 6h, and abdominal distension at 1h. Moreover, the percentage of patients without pain was also decreased, although the number of included RCTs was less. Whether CO 2 insufflation induces metabolic disorder such as CO 2 retention and decrease in SpO 2 % remains to be clarified. Actually, all included trials in this meta-analysis showed concordance between the two groups, and the merged results for all individual studies also revealed no significant difference. In addition to CO 2 insufflation, respiratory depression caused by conscious sedation may also lead to CO 2 retention. We also presented the dosage of sedative drugs used. Only Saito et al [15] reported a difference when patients were administered midazolam alone; for the other studies [13,14,20,22] there ws no significant between-group difference with respect to dosage of sedative drugs. Moreover, a significant heterogeneity was observed among the studies owing to the use of combination of sedative drugs and the differences with respect to patient's sedation level.
Until now, two studies [7,18] have analyzed residual gas; the results bore some similarities to abdominal discomfort caused by rapid absorption into the bloodstream. The study by Takada et al [23] was the first RCT to show that use of CO 2 insufflation for ESD could reduce the risk of Mallory-Weiss tear [14]; similar results were later reported by Onogi et al [17]. The lower incidence of Mallory-Weiss tear in the CO 2 group was ostensibly due to lower tension of the gastric mucosa as a result of residual gas in the stomach. More RCTs are needed to assess this issue. Four complications were reported in the seven studies included in the present metaanalysis [14,15,[19][20][21][22]. When perforation occurs, the gas in the gastrointestinal tract leaks into the peritoneal cavity. Owing to faster absorption of CO 2 , use of CO 2 insufflation is expected to lead to lower intra-abdominal pressure in the postoperative period; however, our meta-analysis showed no significant between-group difference in this respect. Contrary to earlier reports, use of CO 2 insufflation was associated with shorter procedure time than that associated with use of air insufflation. However, more than half of the included studies did not report the experience level of the surgeons, which could have affected the procedure time.

Limitations
This meta-analysis has several limitations. First, all included studies were from Asia. Second, several studies excluded patients with severe pulmonary disease. Although Takada et al [23] recently reported the safety of CO 2 insufflation during ESD in patients with pulmonary Table 3. Changes in the heterogeneity of abdominal pain VAS score at 24h post-procedure after sequential exclusion of one study at a time. dysfunction under conscious sedation, it was a single-center and uncontrolled study. Other notable limitations of our meta-analysis include heterogeneity among the included studies, inconsistency in the reported outcomes, incomplete data on some of the variables, the relatively small number of included studies and the possibility of publication bias. Some of the outcomes such as abdominal distension are liable to be effected by the depth of sedation and the skills of the endoscopist. Subgroup analyses to assess the effect of these variables could not be performed due to lack of original data. Only two studies reported data on abdominal distension [13,14]. More robust studies are required to draw definitive conclusions.

Conclusion
In conclusion, this meta-analysis indicates that CO 2 insufflation may offer advantages over air insufflation with respect to postoperative abdominal discomfort, procedure time, and the residual gas volume. No significant difference was observed with respect to PtcCO 2 , PaCO 2 , SpO 2 %, abdominal circumference, length of hospital stay, sedation dosage, complications, WBC count, serum CRP level, and dysphagia. More RCTs are required to assess its advantages and necessity in future.
Supporting information S1