Comparative Efficacy of Rectal Nonsteroidal Anti-Inflammatory Drugs and Pancreatic Duct Stents in Preventing Pancreatitis After Endoscopic Retrograde Cholangiopancreatography: A Network Analysis Cholangiopancreatography: A Network Analysis.

Background/Aims: Some meta-analyses suggest that pancreatic duct stents (PDS) placement and rectal nonsteroidal anti-inflammatory drugs (NSAIDs) can prevent post-endoscopic retrograde cholangiopancreatography pancreatitis (PEP), but which is favorable is unclear. Thus, we conducted network analyses to compare treatment efficacies. Materials and Methods: The Cochrane controlled trial registry, PubMed, China Biology Medicine disc (CBM disc) and Embase were reviewed, focusing on the randomized controlled trials (RCTs) of rectal NSAID and PDS prophylaxis throughout the PEP prophylaxis process On September 25, 2019. Reviewers extract data independently and randomly. The Cochrane handbook was used as a standard to assess migration risk, and intervention management was implemented in a systematic way. The rationality of the evidence is determined by means of grade evaluation, proposal formulation and evaluation, and the main output result is the effect of intervention. Results: OR= 0.38 95% CI 0.19 to 0.78. Network analysis of all trials demonstrated the quality of PDS OR 0.32 95% CI 0.21 to 0.48, DIC: OR 0.30 95% CI 0.18 to 0.49, ID: OR= 0.55 95% CI 0.39 to 0.77, NAP: OR= 0.55 95% CI 0.39 to 0.77. Network analysis ranking showed that PDS placement was the best way for preventing PEP, followed by NAP, ID, and DIC. Conclusion: This evidence suggests that PDS placement is most effective in reducing PEP risk. However, rectal NSAIDs may be more ideal due to decreased costs.

Drug prophylaxis of PEP has always been a hot topic in clinical research, the optimal drugs for preventing PEP have common characteristics such as efficacy, low side effect profile, high costbenefit ratio, easy acquisition and easy administration. For the people with the process and high-risk patients may be at higher risk, NSAID for preventing the patient produces PEP has the extremely widespread application value, some meta-analysis [6][7][8] by pointed out that such as the European association for gastrointestinal endoscopy and Japan GanDanYi surgical association guidelines are referred to such a question, namely for high-risk patients can be treated with NSAIDs [9][10][11][12][13]. There are also a number of metaanalyses that show that PEP can be largely prevented by stenting the pancreas. But the two approaches also raise questions about whether the interventions are theoretically complementary and whether they are synergistic. We need to give further demonstration and research. Network meta-analysis was used to effectively analyze the data. It includes studies on various types of treatment and prevention strategies that show that rectal maids are significantly more effective than pancreatic stents in the prevention of PEP [14], but a few studies have a publishing bias against these findings, non-randomized controlled trials. Thus, the main purpose of this study is to compare the differences between the two through network meta-analysis and system review NSAIDS and stents for preventing PEP, all randomized controlled trials were involved and prevention was performed for the best PEP intervention The RCTrelated data were searched comprehensively. In order to eliminate language restrictions, we generally use the following terms: non-steroidal anti-inflammatory drugs, pancreatic*, endoscopic retrograde cholangiopancreatography, ERCP, NSAIDS, stent, random* combined with all content included rectal NSAIDs. We included RCTs that compared rectal NSAIDS or stents to placebo controls or other rectal NSAIDS, used for the prevention of PEP in patients. We analyzed only RCTs because placebo was included as part of the network meta-analysis, for RCT study design, it is mainly used to ensure the rigor and accuracy of the whole study process by determining the deviation and minimizing the performance [15]. Two reviewers, Zhang Qian and wang Yin, reviewed the search results to prevent the reports from being ignored, if there were any disagreements, the third reviewer (Fu-Cheng Zhang) made the final decision.

Data extraction
The uniformity of the data extracted for the purpose of lifting, the reviewers, (Yin Wang, Qian Zhang) independently extracted the original data presented in the literature onto a standardized form: country, inclusion criteria, year of publication, intervention time, the first author, Rectal NSAIDs (drug, dose, NO.), control (placebo (PLA)/NO.), NO. of PEP. If necessary, the author of the study was contacted in order to obtain the study data. The conflict of abstract data can be solved by negotiation and reference to the original article.

Ethic committee approval
Because the type of the paper is reviewing whose materials are derived from primitive clinical research from other authors, we do not have Informed Consent necessarily

Risk of bias assessment
The authors mainly used the Cochrane collaboration manual [16] as a reference for independent evaluation of the quality of the literature. The criteria included in the scoring system are selective scaling, ignorance of assessment results, incomplete result data, and ignorance of participants, random sequence generation and other sources of bias. The results diagrams presented were created by using the soft Reman 5.2 (The Nordic Cochrane Center, Denmark).

Statistical analysis
Stata15.0 (StataCorp, Texas) was used in our analysis for standard meta-analysis to achieve direct comparison. Two different types of data were used to study and analyze the likelihood of the onset of PEP, and to calculate the confidence curve and relative risk, and effectors. The heterogeneity between the results of the study were examined using the Q test (test level is a=0.1), and the magnitude of heterogeneity was judged by combining the findings with the I2 test [17]. I2-related studies with volume fractions in the 25% to 50% range had low heterogeneity. At the same time, I2-related studies with volume fractions ranging from 50% to 75% had moderate heterogeneity. Studies of I2 with a volume fraction of more than 75% had higher heterogeneity. If the I2 score is more than 50%, then the model should be constructed by random effect.
If not, then a fixed effect model needs to be built for the above situation. Dual-sample test and funnel plot were used to evaluate the deviation. In addition, dual-sample test was used to evaluate the heterogeneity between target objects. If P <0.05, it means that there is a significant difference in the field of statistics. At the same time, the possibility of publication deviation can also be evaluated by means of funnel plot asymmetry. Refer to PRISMA's statement for meta-analysis.

Network meta-analysis
When carrying out data analysis, we mainly used network meta-analysis with the software Stata15.0 and the momenta command. The first step is to build a network of evidence whose main purpose is to compare treatments. By means of 95%Cl and inconsistent factors, the consistency of each closed-loop part was evaluated and analyzed. The consensus is that 95%Cl is 0. If not, then this part of the closed loop will be identified as significantly inconsistent. In this paper, all output indexes are used for data counting. Therefore, RR combination is used to combine the effect, and 95%Cl is used to estimate the interval size. The upper limit of 95%Cl mentioned above is less than 1, while the lower limit is greater than 1, according to the above results, it can be known that there are significant differences between the two, otherwise it would be considered that there was no significant difference in

GMR.000593. 4(4).2020
the field of statistics. Results were obtained after a network metaanalysis, and PR Best and SUCRA were used to rank the results. For subgroup analysis, higher SUCRA scores were associated with better therapeutic outcomes. In addition, PR Best represents the probability that the treatment is the best treatment. After obtaining the data results, a recommendation assessment, a development assessment, and a registration assessment [18] are required for each result, and the specific results obtained are displayed in each breakdown. If there is no deviation limitation due to reporting deviation, research treatment, inconsistency, data sparsity, inaccuracy, etc., then the obtained RCT will be considered as highquality certification. If any of the above occurs, the level of evidence will be classified as moderate or low.

Copyright © Yin Wang
GMR.000593. 4(4).2020 33,35,36], diclofenac in 8 studies [19,20,25,26,[28][29][30]37], naproxen in 2 studies [19,34], and stents in 13 studies [38][39][40][41][42][43][44][45][46][47][48][49][50]. Sample sizes for the studies ranged from 19 to 665. The outcomes contain severe pancreatitis, moderate, incidence, and mild. Table 1 & 2 below specifically present all the features included in this network meta-analysis. Wang Yin and Zhang Qian mainly used the Cochrane collaboration tools to conduct migration risk assessment studies and conducted evaluation reviews for the quality of the methods they used. Each experiment was evaluated as a whole, including high risk, unknown risk and low risk. Among them, the reviewers shall deal with and solve the differences existing in the quality assessment. (Fu-Cheng Zhang). The summary of performance bias, reporting bias, attrition bias, selection bias and other biases identified for each individual RCT is shown in Figure 2a & 2b. All the results get from controlled trials showed that the migration risk was unclear or low.

Standard meta-analysis
Combined with 18 comparative experiments, through the comparative analysis between placebo and rectal NSAIDS, among all the patients participating in the study, the participants in the control group and the intervention group were 2,542 and 2,507, respectively. At the same time, the random effect model was constructed for the analysis objects. It was known that the heterogeneity I2 between all the subjects participating in the study was 44.5%. A strong association was found between the treatment group and the incidence of PEP when compared with placebo (OR= 0.46, 95% CI 0.34 to 0.61, I2=44.5%). When we divided these trials into three subgroups based on the type of administration, intervention time, and the inclusion criteria, the type of administration indicated that the probability of occurrence of PEP showed a significant downward trend，DIC OR= 0.31 95% CI 0.16 to 0.59 I2=47.3%; ID OR= 0.56 95% CI 0.41 to 0.76 I2=33.8%; NAP OR= 0.38 95% CI 0.19 to 0.78. For intervention time, three subgroups -before, after and during ERCP -were analyzed, before ERCP OR= 0.44 95% CI 0.34 to 0.57 I2=7.2%; after ERCP OR= 0.41 95% CI 0.19 to 0.88 I2=61.1%; during ERCP OR= 1.51 95% CI 0.68 to 3.33. As for the standard of the inclusion criteria, two subgroups with high risk patients and unselected patients were analyzed, with high risk patients, OR= 0.37 95% CI 0.20 to 0.69 I2=57.8%; with unselected patients, OR= 0.51 95% CI 0.37 to 0.71 I2=36.1%. The various subgroup analyses carried out revealed a consistency between its results and the overall results, summarized in Figure  2c-2e. According to the stent placement vs placebo in the 13 trials, the total number of participants in the treatment trial was 827 in In order to compare and rank rectal NSAIDS and stent placement based on their ability to prevent PEP, the RCT data were taken as the target object and the network meta-analysis was carried out. According to the findings from the network meta-analysis using all trials, DIC (OR 0.30, 95% CI 0.18-0.49), ID (OR 0.52, 95% CI 0.37-0.74), NAP (OR 0.78, 95% CI 0.37-1.64), PDS (OR 0.32, 95% CI 0.21-0.48), DIC, ID, and PDS were all significantly associated with the reduction or prevention of PEP, whereas NAP was not. The results of network-analysis for preventing the incidence of PEP from all the trials is given in Table 3. Surface under the cumulative ranking curves (SUCRA curves) for preventing the incidence of PEP are presented in Figure 3e.

Discussion
This network meta-analysis included all appropriate administrations often used by clinicians for preventing PEP, such as rectal NSAIDS or stent placement. The migration risk associated with all experiments is unclear or low. The results indicate that rectal NSAIDs and PDS significantly decrease the incidence of PEP when compared to placebo from the pairwise meta-analytic results. DIC and ID use demonstrate high quality evidence, there is a low level of quality for NAP related evidence with evidence from only one trial. Moderate heterogeneity can be seen with rectal diclofenac, in the trial from Lua G [25], it shows that there is no statistically significant difference between PLA and DIC. According to the results of the study on the use of NSAIDS in the rectum, the possibility of the occurrence of PEP can be greatly reduced, therefore, it is clear that the standard results are the same as the experimental results. For PDS, the incidence of PEP has been significantly reduced according to the results of some high-quality evidence.
In order to further improve the accuracy of the results we have obtained, so as to support clinical purposes, we now focus on further strengthening face-to-face studies and improving the credibility and quality of evidence. Our assessment overall found little difference between NSAIDs and PDS when all data were considered, except for NAP which had only one direct comparison, a network meta-analysis showed a lower level of quality. The remaining results were the same as the standard meta-analysis. At the same time, this ranking can indicate that the placement of stents is the best option among various ways to prevent PEP There was only one network meta-analysis comparing the rectal NSAIDs to stents.
We also make sure that the content is as comprehensive as possible when conducting literature search, including all published RCTs. Our research has several advantages as follows: first, we have expanded the study populations significantly. Second, this kind of network meta-analysis mainly includes the majority of which are high quality, however, the previous network meta-analysis included many retrospective experiments which may increase the heterogeneity and the credibility of results. Last, the present results indicate that placement of PDS was more effective for PEP prevention than the NSAIDs. There is clearly a significant difference between the results of this study and those obtained by conducting Copyright © Yin Wang GMR.000593. 4(4).2020 the meta-analysis described above. We first introduced the GRADE system to assess our results, which may be more helpful to clinicians, and compared the different types of rectal NSAIDs, and determined which type is the best choice for preventing PEP. Despite the above, we also have limitations in our experimental work that need to be addressed. Heterogeneities in some analyses including direct and mixed comparisons were moderate. According to the standard meta-analysis, we found that the heterogeneities were derived from the different types of NSAIDs. With the mixed comparisons, some loops were inconsistent, and through the funnel plot, we can tell that all PDS have the possibility of publication deviation, all the inconsistence and probable publication biases may be due to the number of direct comparisons among the rectal DIC, ID, NAP, and some small sample trials with PDS. Although the statistics suggest that placement of PEP stents were superior to rectal NSAIDs, this may not be ideal for daily clinical operations due to the complicated procedure, risk associated with the intervention, excess expense and inconvenience to the participants. Rectal NSAIDs may be the cheapest and most convenient choice, ESGE protocol recommends the use of rectal NASIDS to prevent PEP. However, the lack of direct comparisons between rectal NSAIDs and direct comparisons between rectal NSAIDs and PDS, suggest the need for further RCTs between stents and rectal NSAIDs to confirm our results. Taken together, stents and rectal NSAID can reduce the likelihood of PEP occurring. The available evidence from network meta-analysis suggests that stents are more effective in reducing the risk of PEP, because of the no significant contrast of rectal NSAID, there may be heterogeneity in our results, therefore, in the process of indirect comparison and direct comparison, high-quality RCT should be used as the basis to enhance the credibility of the results.

Author's Contribution
Yin Wang: data collection, data extraction, data analysis and writing the article.
Qian Zhang: data collection, data extraction, data analysis and modifying the article.
Fu-Cheng Zhang: Technology and method guidance, handle differences.