Inhibition of Aspirin-Induced Gastrointestinal Injury: Systematic Review and Network Meta-Analysis

Background: Administration of aspirin has the potential for significant side effects of gastrointestinal (GI) injury mainly caused by gastric acid stimulation, especially in long-term users or users with original gastrointestinal diseases. The debate on the optimal treatment of aspirin-induced gastrointestinal injury is ongoing. We aimed to compare and rank the different treatments for aspirin-induced gastrointestinal injury based on current evidence. Methods: We searched PubMed, EMBASE, Cochrane Library (Cochrane Central Register of Controlled Trials), and Chinese databases for published randomized controlled trials (RCTs) of different treatments for aspirin-induced gastrointestinal injury from inception to 1 May 2021. All of the direct and indirect evidence included was rated by network meta-analysis under a Bayesian framework. Results: A total of 10 RCTs, which comprised 503 participants, were included in the analysis. The overall quality of evidence was rated as moderate to high. Eleven different treatments, including omeprazole, lansoprazole, rabeprazole, famotidine, geranylgeranylacetone, misoprostol, ranitidine bismuth citrate, chili, phosphatidylcholine complex, omeprazole plus rebamipide, and placebo, were evaluated in terms of preventing gastrointestinal injury. It was suggested that omeprazole plus rebamipide outperformed other treatments, whereas geranylgeranylacetone and placebo were among the least treatments. Conclusion: This is the first systematic review and network meta-analysis of different treatments for aspirin-induced gastrointestinal injury. Our study suggested that omeprazole plus rebamipide might be considered the best option to treat aspirin-induced gastrointestinal injury. More multicenter, high quality, large sample size randomized controlled trials will confirm the advantages of these medicines in the treatment of aspirin-induced gastrointestinal injury in the future.


INTRODUCTION
Aspirin is a well-known baseline antiplatelet agent, serving as the primary and secondary prevention as well as the treatment of ischemic stroke, acute myocardic infarction, transient ischemic attack, acute coronary syndrome, and peripheral artery diseases (A. J. Hermosillo and Spinler, 2008). However, many studies have shown that the administration of aspirin has the potential for significant side effects of gastrointestinal (GI) symptoms including mucosal lesions, bleeding or peptic ulcers, especially in long-term users or users with origin gastrointestinal diseases (R. Guthrie, 2011). The aspirin-induced gastric injury is a chronic inflammation mainly caused by gastric acid stimulation, as aspirin break the gastric mucosal barrier; gastric acid is capable of breaking the mucosal epithelial cells directly, causing inflammation, bleeding, and gastric ulcer. However, the mechanism of aspirin-induced small bowel injury has remained incompletely understood; its occurrence involves intestinal microorganism, bile and other stimulating factors, which is more complex than the gastric injury caused by aspirin (Washio, et al., 2016;Wallace, 2012). Therefore, there are no guideline recommendations for the treatment of aspirininduced gastrointestinal injury.
Medications such as proton pump inhibitors (PPIs), histamine 2 receptor antagonists (H2RAs), misoprostol, and alendronate are used clinically to prevent aspirin-induced gastrointestinal injury. Recent randomized controlled trials (RCTs) have demonstrated that these treatments have high therapeutic potential effects for aspirin-induced gastrointestinal injury. However, it is difficult to directly rank the efficacy and safety of the treatments due to the limitations of study design and scale.
Comparing to conventional meta-analysis, network metaanalysis (NMA) offers more advantages, which has been increasingly advocated in medical research, as it provides more comprehensive analytical evidence for selecting the optimal treatment by directly and indirectly evaluating different intervention models (Nikolakopoulou et al., 2018). To address this knowledge gap, we conducted a systematic review and network meta-analysis of randomized controlled trials (RCTs) of all the medications used in aspirin-induced gastrointestinal injury to evaluate their relative efficacy and safety.

METHODS
Our systematic review and NMA manuscript was written in accordance with the Cochrane Handbook (Higgins et al., 2020). Reporting was consistent with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) extension statement for reporting systematic reviews incorporating NMA (Guise et al., 2017).

Data Sources and Searches
We searched PubMed, EMBASE, Cochrane Library (Cochrane Central Register of Controlled Trials), and two Chinese databases (China National Knowledge Internet database and China Wanfang database) for citations published in any language from inception to 1 May 2021. We used combinations of MeSH terms and text words around "aspirin," "acetylsalicylic acid," "gastrointestinal injury," "gastrointestinal lesions," "peptic ulcer," "bleeding," "perforation," "obstruction," "healthy people," "healthy subjects," "randomized controlled trial," and "clinical trial." Additional studies were derived from screening the reference lists of included RCTs and previous systematic reviews. We also searched trial registers like ClinicalTrials.gov, iSCTRN, and OpenTrials.net to identify any unpublished or ongoing trials and contacted the researchers to ask about unpublished studies.

Eligibility Criteria
We included the RCTs with the following criteria: 1) RCTs inhibiting aspirin-induced gastrointestinal injury with medication in healthy people; 2) RCTs reporting number of people with gastrointestinal lesions, with or without safety outcome; 3) minimum sample size no less than 20.
We excluded the RCTs with the following criteria: 1) RCTs with significant flaws, incomplete or wrong results; 2) dissertations and conference abstracts; and 3) RCTs with the participant numbers in the smaller group less than 20. Besides, we excluded observational, cross-sectional, case series, or qualitative studies.

Data Extraction and Quality Assessment
Two independent authors (Wan-tong Zhang and Bao-chen Zhu) assessed all RCTs for eligibility and extracted data and assessed study quality. Disagreements were resolved by consensus-based discussion. The full-text articles were downloaded and the same inclusion criteria were used to decide whether to include or exclude at this stage. The average age and sex ratio of patients, grouping, intervention measures and outcome indicators were recorded. Each reviewer independently carried out the quality assessment of each selected article, assessed the completeness of the data extraction, and confirmed the quality rating to reduce bias.
Three authors (Wan-tong Zhang, Bao-chen Zhu, and Miaoran Wang) evaluated all the included RCTs. Any disagreement was solved through discussion, mutual consensus, and rechecking of the article. For the RCTs which were included, the Cochrane risk-of-bias tool was used to assess risk of bias (Sterne et al., 2019), including selection bias (random sequence generation/allocation concealment), performance bias (blinding of participants and personnel), detection bias (blinding of outcome assessment), attribution bias (incomplete outcome data), reporting bias (selective outcome reporting), and other possible sources of bias. In this way, all articles selected for inclusion in the review were graded under the categories of low, high, or unclear risk of bias.
The efficacy outcome was the number of individuals diagnosed with gastrointestinal injury after taking aspirin. Safety assessments included adverse events (AEs).

Data Synthesis
For further evaluation of gastrointestinal protecting medications simultaneously, a network meta-analysis was used to synthesize Frontiers in Pharmacology | www.frontiersin.org August 2021 | Volume 12 | Article 730681 the study effect sizes (both direct and indirect comparisons). Heterogeneity across individual studies was estimated by the Cochran Q test (Chi-squared) and Higgins I-squared inconsistency statistic. If there was no significant heterogeneity (p > 0.05, I-squared <50%), a fixed-effects model was used. Otherwise, a random-effects model was used. Moreover, the analysis of ranking probabilities and the Surface under the Cumulative Ranking curve (SUCRA) were used to explore the probabilities of each treatment being ranked the best. The SUCRA value, expressed as a percentage, is one of the best choices for indicating the relative probability of intervention. Inconsistency in the network was evaluated using the deviance information criterion (DIC) between the consistent and inconsistent models. If the DIC difference was within 3, the data were generally considered to be consistent. Once considerable heterogeneity (p < 0.05, I-squared>50%) was found in pairwise meta-analyses, we will conduct a sensitivity analysis; that is, we will remove those studies that lead to high heterogeneity indirect comparisons or introduced statistical inconsistency in the NMA. All data were analyzed by using the R 3.6.3 software and Review Manager version 5.4.

Study Screening
By searching five different electronic databases and manual reference searching, 241 relevant publications were found after removing duplications and unrelated studies. Relevant articles were further selected by reading the full texts. By excluding publications that contained noneligible data, had unsuitable comparator arms, and were not clinical trials and nonrandomized trials, 10 articles were included. Figure 1 shows the detailed steps of the literature selection process.

Results of Meta-Analyses
We further ranked all treatments according to SUCRA; Figure 6 presented all treatments ordered by their probability to be the best treatment in terms of gastrointestinal injury. It was suggested that omeprazole plus rebamipide, ranitidine bismuth citrate, and omeprazole outperformed other treatments, whereas geranylgeranylacetone and placebo were among the least treatments.
The inconsistency in the network for the nine interventions was summarized under consistency and inconsistency assumptions (Supplementary Tables); the difference of DIC  between two models was no more than three, which means the results are generally considered consistent. In the heterogeneity analysis, global I-squared did not identified any heterogeneity across the studies (Supplementary Tables; global I 2 16.21).

DISCUSSION
Aspirin-induced gastrointestinal damage limits its regular use. PPI, H2RAs, misoprostol, alendronate, and some other gastrointestinal protecting medications are widely used in clinical practice for preventing the injury; however, there are no guideline recommendations for the injury. Therefore, identifying the treatment with the best safety and efficacy is important. To the best of our knowledge, there is currently no comparison of multiple interventions for treating aspirin-induced gastrointestinal injury; this is the first systematic review and network meta-analysis investigating different treatments for aspirin-induced gastrointestinal injury. Our study suggested that the combination of omeprazole and rebamipide was potentially the preferred intervention. The mechanism of aspirin-induced gastrointestinal injury is usually considered to be divided into two parts: direct injury and indirect injury. The indirect mechanism of damage to the gastrointestinal mucosa by aspirin is the nonreversible inactivation of cyclooxygenase (COX), which inhibits the conversion of arachidonic acid to prostaglandin H2 (PGH2) in platelets (Hart and Bailey, 2002), subsequently affecting the production of prostaglandins PGE2 and PGI2. The decreased secretion of PGE2 decreases gastric mucus and HCO3-secretion and reduces gastrointestinal blood flow (Takeuchi, et al., 2015). The decreased secretion of PGI2 also decreases gastrointestinal lower blood flow (Harada et al., 1999). The limited energy caused by the lower blood flow reserved in the gastrointestinal mucosa makes it very sensitive to ischemia and hypoxia. It might cause damage to epithelial cells of the gastrointestinal mucosa, reduce FIGURE 4 | Efficacy of all interventions according to network meta-analysis. Drugs are reported in order according to efficacy ranking. Treatment at the top left corner ranks first, while the one at the bottom right corner ranks last. Comparisons between treatments should be read from left to right and the estimate is in the cell common between the column-defining treatment and the row-defining treatment. ORs (95% CrI) lower than one favor the column-defining treatment. Results are rounded to two decimal points. Significant results are in bold.

FIGURE 5 | Direct meta-analysis.
Frontiers in Pharmacology | www.frontiersin.org August 2021 | Volume 12 | Article 730681 6 mucosal barrier function, increase permeability and the susceptibility to endotoxin and bacterial attack. Another indirect mechanism of damage is that aspirin could affect intestinal flora, in particular, Gram-negative bacteria. It stimulates the immune system characterized by Toll-like receptor 4 and finally induces inflammation (Otani et al., 2017). The main ways in which aspirin causes direct gastrointestinal damage include 1) disrupting the hydrophobicity of mucus HCO 3 − protective layer of the gastric mucosa (Goddard and Hills, 1987); 2) damaging the gastrointestinal mucosal epithelial cells, including inducing oxidative stress damage to the gastrointestinal mucosal epithelial cells, elevating malondialdehyde concentration and inducing apoptosis in the gastrointestinal mucosal epithelial cells (Hernandez et al., 2016); and 3) disrupting the permeability of the gastrointestinal tract and causing inflammation (Lai et al., 2015). In summary, aspirin-induced gastrointestinal damage may be caused by a combination of inhibition of mucus secretion, inhibition of mucosal blood flow, oxidative stress damage, and inflammation, which is observed in our previous preclinical research (Zhu, et al., 2018). In such cases, inhibition of gastric acid production alone may not fundamentally inhibit aspirin-induced gastrointestinal injury. The treatment should be considered to elevate mucosal blood flow, reduce inflammatory infiltration, and promote mucus production at meantime. In animal experiments, rebamipide inhibited the infiltration of inflammatory cells in the gastric mucosa and increased the amount of gastric mucus, gastric mucosal blood flow, and gastric mucosal prostaglandin content. However, rebamipide showed no inhibitory effect on stimulating gastric acid secretion (Arakawa et al., 2005). Therefore, the combination of rebamipide with PPIs whose main effect is to inhibit gastric acid secretion can mechanistically inhibit aspirininduced gastrointestinal injury.
H2RA and PPI are also used in clinical practice against aspirin-induced gastrointestinal injury. The results suggest that ranitidine bismuth citrate, omeprazole, and lansoprazole showed greater effectiveness than placebo. Ranitidine bismuth citrate is a combination drug of H2RA with bismuth citrate. The gastroprotective effects of ranitidine bismuth citrate include inhibiting gastric acid secretion (Tanaka, et al., 1996), inhibition of pepsin activity (Tay, et al., 1990), binding to ulcers (Koo et al., 1982), stimulation of prostaglandin synthesis and bicarbonate secretion (Konturek et al., 1987;Mertz-Nielsen et al., 1991), antibacterial activity (Manhart, 1990), and binding to mucus (Lambert, 1991). PPIs could inhibit gastric acid secretion and pepsin activity (El et al., 2018). Although not as good as the combination of rebamipide and PPIs, the use of PPI and ranitidine bismuth citrate alone can also be effective clinically.
In addition to the efficacy, safety is also an important issue in the treatment of aspirin-induced gastrointestinal injury. Clinicians have a responsibility to ensure that subjects were informed about the potential adverse drug reactions. The results showed that the combination of aspirin and protective treatments for gastrointestinal injury showed less side effects. Adverse events just happened in two treatments which are diarrhoea, dyspepsia, headache, aphthous ulcer, tinnitus, nausea, heartburn, abdominal pain, skin rash, and facial puffiness.
Compared with previous conventional pairwise metaanalyses, several merits of this study need to be highlighted. First, this study used network meta-analysis to explore the relative benefits of different treatments for aspirin-induced gastrointestinal injury and inform related clinical guidelines. The Bayesian model used in our study is currently the most applicable approach for multiple-intervention network metaanalysis, which could address the absence of direct comparison and reveal a favorable intervention by ranking analysis. Second, our literature search was as comprehensive as possible, which not only retrieved electronic databases commonly used in Chinese and English but also retrieved references from the reference lists of included RCTs and previous systematic reviews. However, this study also has some limitations, such as the uneven chronological distribution of the included literature, while the doses of aspirin used in some 1990s literature are too large compared to the current treatment regimen. As a result, this comprehensive search strategy for the retrieval of the maximum number of available published trials and a predesigned inclusion criterion ensured the lowest possible degree of heterogeneity. There are several implications and considerations related to these findings. First, this study was conducted in healthy subjects to provide clinical indications for subsequent long-term users of aspirin. Second, this paper includes a variety of drugs that may play a role in prevention and treatment of aspirin-induced gastrointestinal injury, including novel drugs, to inform future clinical studies and guideline development.

CONCLUSION
This is the first systematic review and network meta-analysis of different treatments for aspirin-induced gastrointestinal injury. The findings from this network meta-analysis represent the most comprehensive analysis of the available evidence. Our findings suggested that omeprazole plus rebamipide might be considered the best option to treat aspirin-induced gastrointestinal injury. These results should be taken into account in future guidelines and the selection of interventions for aspirin-induced gastrointestinal injury. Besides, more multicenter, high quality, large sample size randomized controlled trials will confirm the advantages of different medicine in the treatment of aspirininduced gastrointestinal injury in the future.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.