A systematic review and meta-analysis

Background: Whether to conduct enteral nutrition in patients with severe acute pancreatitis (SAP) during the active phase of intestinal stress or to feed during remission remains controversial. This study was aimed to evaluate the efficacy and safety of enteral nutrition within 48hours after admission in the patients with SAP or predicted severe acute pancreatitis (pSAP). Methods: We searched PubMed, EMBASE, Web of Science, and the Cochrane Library before December 2017. Randomized controlled trials of early enteral nutrition (starting within 48hours after admission) versus late enteral nutrition or total parental nutrition in severe acute pancreatitis or predicted severe acute pancreatitis were selected. Results: Ten randomized controlled trials containing 1051 patients were included. Comparing early enteral nutrition to late enteral nutrition or total parental nutrition in SAP or pSAP, the pooled risk ratios were 0.53 (95% confidence interval [CI] 0.35–0.81, P= .003) for mortality, 0.58 (95% CI 0.43–0.77, P= .0002) for multiple organ failure (MOF), 0.50 (95% CI 0.33–0.75, P= .0008) for operative intervention, 0.75 (95% CI 0.61–0.93, P= .009) for systemic infection, 0.42 (95% CI 0.26–0.69, P= .0005) for local septic complications, 0.84 (95% CI 0.74–0.96, P= .01) for gastrointestinal symptoms. 0.87 (95% CI 0.74–1.02, P= .08) for systemic inflammatory response syndrome (SIRS), and 1.24 (95% CI 0.66–2.31, P= .50) for other local complications. Conclusions: Enteral nutrition within 48hours after admission is efficient and safe for the patients with SAP or pSAP. Abbreviations: MOF =multiple organ failure, pSAP= predicted severe acute pancreatitis, SAP = severe acute pancreatitis, SIRS = systemic inflammatory response syndrome.


Introduction
Severe acute pancreatitis (SAP) is one of the devastating diseases leading to intensive care unit (ICU) admission. [1] The high mortality is due to the serious complications including systemic inflammatory response syndrome (SIRS) and multiple organ failure (MOF). [2] Animal and human studies [3,4] have demon-strated that the damage of intestinal barrier function accelerates the development of local and systemic infectious complications. In the early stage of SAP, the intestinal permeability has been significantly increased, which results in the translocation of inflammatory mediators and toxic products. [5] Furthermore, the gut microbiota gets the chance to the systemic circulation through the damaged intestinal epithelial cells. As a consequence, the sepsis or infected pancreatic necrosis occur in the early stage of SAP. [6] So, the maintenance of intestinal barrier function in the early stage is critical for the mortality and prognosis.
In SAP, the inflammatory response induced by necrosis or secondary infection leads to the increase of caloric requirement and loss of mass protein. This contributes to the nutritional deterioration and negative nitrogen balance, which further lead to the damage of function and structure of vital organs. [7] So, early nutritional management plays an important role in the patients with SAP. Parenteral nutrition (PN) has been regarded as the standard care for providing nutrients and can avoid pancreatic stimulation. But, PN may lead to intestinal atrophy and attenuate the intestinal barrier function. [8] Enteral nutrition (EN) is found to be better at maintaining the function and structure of intestinal mucosa. [9] For avoiding the pancreatic stimulation, the route of EN is delivered beyond the Treitz' ligament, which results in minimal or negligible stimulation. [10] Some reviews have suggested that early EN was associated with a reduction in mortality, MOF and infections compared with the late EN or PN in the patients with acute pancreatitis. [11][12][13] But the above studies were not strictly stratified on the basis of the severity of disease, which potentially led to selection bias and may misguide the therapy of SAP. Despite these data, whether to conduct EN in patients with SAP during the active phase of intestinal stress or to feed during remission remains controversial. Recent trials have yielded mixed results and the guidelines are vague and contradictory. [14][15][16] Therefore, we attempt to rigorously design and comprehensively reevaluate the efficacy and safety of early EN (starting within 48 hours after hospital admission) in the patients with SAP or predicted SAP.

Search strategy
This meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. [17] We independently searched PubMed, EMBASE, Web of Science, and the Cochrane Library before December 2017 for relevant studies. The detailed search strategy was in the appendix (Supplemental Digital Content, http://links.lww.com/MD/C404, which illustrates the detailed search strategy). No restrictions were placed on language.

Selecting criteria
Studies included in this meta-analysis fulfilled the following criteria:

Types of outcome measures
The clinical outcomes are the following: (1) mortality;

Data extraction and management
Two independent reviewers used a standard form for data abstraction. The extracted data were cross-checked by the reviewers. The extracted information included: first author, publication year, country of origin, study design, patient demographics, sample size, type of intervention, and outcomes.
Disagreements between reviewers were resolved by the third reviewer.

Assessment of risk of bias in included studies
The quality of the included randomized clinical trials (RCTs) was assessed according to the methodological criteria of the Cochrane Handbook for Systematic Reviews of Interventions. We assessed the risk of bias through seven domains, including allocation sequence generation, allocation concealment, blinding of participants and study personnel, blinding of outcome assessors, management of incomplete outcome data, selective outcome reporting, and other potential sources of bias. The publication bias was evaluated by funnel plots if ten or more studies were included in an outcome. Grading of Recommendations Assessment, Development and Evaluation (GRADE) system was used to create a summary of findings table and assess the quality of evidence. [18] 2.6. Statistical analyses All statistical analyses were performed using RevMan Software (version 5.3) and STATA software (version 12.0). Binary variables were combined to estimate the pooled risk ratio (RR) with 95% confidence intervals (CIs). The I 2 test was used to measure statistical heterogeneity among the included studies and P < .1 or I 2 > 50% indicated significant heterogeneity. If significant heterogeneity was not observed, a fixed-effects model was used to make estimates, otherwise, a random-effects model was applied to statistical analysis. A P value < .05 was considered statistically significant.

Search results
A total of 1424 articles were found from PubMed, EMBASE, Web of Science, and the Cochrane Library. The flow diagram for searching and screening of eligible studies is shown in Figure 1. Finally, 10 RCTs including enrolled 1051 patients were included in this meta-analysis.

Characteristics of trials included
We included ten randomized controlled trials of published studies. The 5 of 10 RCTs were the patients with predicted SAP. Predicted SAP was defined as Acute Physiology and Chronic Health Evaluation (APACHE) II score ≥ 8 or C-reactive protein (CRP) levels ≥ 150 mg/L. Countries of publication were diversity, three from China, [19][20][21] one from the Netherlands, [22] Sweden, [23] the UK, [24] Greece, [25] Russia, [26] Poland, [27] Croatia, [28] and all were in English. All included studies were full-text papers. More details of the included studies were illustrated in Table 1.

Risk of bias in included studies
Among 10 included RCTs, 6 studies provided complete data in allocation sequence generation. Four studies did not specifically describe the method of allocation sequence generation. Allocation concealment was adequate in 9 studies. Only 1 study did not provide enough information regarding the use of allocation concealment method. None of the studies provided enough  information regarding the use of the blinding method (Figs. 2 and  3). No more than 10 studies were included in an outcome, so the publication bias was not evaluated by funnel plots. If the asymmetry were obvious in this minimum number of studies, the assessment of funnel plot asymmetry would have low power to differentiate real asymmetry. [29] 3.4. Effects of interventions 3.4.1. Mortality. This outcome was reported in nine studies with 969 patients. The results were homogenous and thus a fixedeffects model was used. After aggregating the data, a significant reduction was observed in the early EN group compared with the late EN or PN group (RR = 0.53, 95% CI 0.35-0.81, P = 0.003, I 2 = 44%). (Fig. 4) 3.4.2. Multiple organ failure (MOF). Eight studies collected data for this outcome with 931 patients. Our results showed that early EN was associated with a significant reduction in the rate of MOF compared with the late EN or PN group (RR = 0.58, 95% CI 0.43-0.77, P = 0.0002, I 2 = 0%) (Fig. 4).

Systemic inflammatory response syndrome (SIRS).
Four studies assessed this outcome including a total of 676 patients. This outcome showed that there was a tendency of decreased SIRS in early EN, but the difference was not significant (RR = 0.87 95% CI 0.74-1.02, P = 0.08, I 2 = 27%) (Fig. 4).

Operative intervention.
This outcome was reported in 7 studies with 643 patients. The results showed that early EN was significantly associated with lower risk of operative intervention than late EN or PN (RR = 0.50, 95% CI 0.33-0.75, P = 0.0008, I 2 = 27%) (Fig. 5).

Other local complications.
Four studies collected data for this outcome with 497 patients. The result showed that there was no significant reduction in the rate of other local complications when comparing the early EN group with the late EN or PN group (RR = 1.24, 95% CI 0.66-2.31, P = 0.50), and the significant heterogeneity was detected (I 2 = 67%, P = 0.03), so a random-effects model was used (Fig. 6).

Summary of findings and quality of evidence
The summary of findings and the GRADE recommendations were illustrated in Table 2.

Discussion
This systematic review evaluated ten RCTs and included 1424 patients with SAP or predicted SAP. The results showed that early EN (starting within 48 hours after admission) significantly reduced the mortality, MOF, operative intervention, systemic infections, local septic complications and gastrointestinal symptoms compared with late EN or PN. In addition, a decreasing trend is found in SIRS, but not significant. Meanwhile, no significant difference is also observed in the risk of other local complications. In GRADE recommendations, the quality of the evidence was moderate for MOF, operative intervention, systemic infections, and local septic complications, low for mortality and SIRS, and very low for other local complications and gastrointestinal symptoms. The intestinal barrier function was described long ago as being involved in SAP, playing an active role in the progression of MOF and infections complications. [30] The early standard EN, even small amounts of EN, may improve the intestinal barrier function through affecting the intestinal permeability, bacterial   translocation, and immunocompetent cells. [31][32][33] Three previous meta-analyses have addressed the issue of early EN (starting within 48 hours after hospital admission). Feng et al [11] included 6 trails with 1007 patients in acute pancreatitis (AP), of which 5 trails with 972 patients were in SAP or predicted SAP. They found that EN within 48 hours was related to a significant reduction in MOF compared to late EN or PN in AP. A decreasing trend is also found in mortality but not significant. We disagree with Feng et al [11] in the fact that they included 2 retrospective studies with 212 participants, which existed the high risk of bias, and they did not make a stratified analysis based on the severity of disease, so the conclusion could not truly reflect the validity of early EN in SAP or predicted SAP. Li et al [12] included eleven trails in AP, of which eight trails with 616 patients were in SAP or predicted SAP. They found a significant reduction in total infection complications, pancreatic infection complications and organ failure in EN within 48 hours compared to the late EN or PN group in SAP or predicted SAP. A significant reduction of mortality was also 418 per 1000 (369 to 478) * The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio; GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. * Moderate statistical heterogeneity. † Some studies were of small sample size. ‡ High statistical heterogeneity.
x The evidence base was at high risk of bias with significant limitations in the performance of blinding. observed, but they did not make a stratified analysis in the indicator of mortality. They also included two retrospective studies with 327 patients in a subgroup of SAP or predicted SAP and these 2 studies were published in abstract form only, which existed great risk of bias. Vaughn et al [34] included eleven trails with 948 patients and indicated that EN within 48 hours did not seem to increase gastrointestinal symptoms, but they mainly focused on patients with mild to moderate AP and had limited data of SAP or predicted. This meta-analysis had a number of limitations. First, some included RCTs were small in size and single center. The blinding was not addressed in all included RCTs, but we acknowledged that the blinding of different feeding routes was impossible, even to radiologists, for example, the nasojejunal tube could be found on radiological images. Second, the feeding routes of EN were different. In our study, the feeding route of one RCT was nasogastric feeding and the others are nasojejunal feeding. But, previous RCT studies indicated that between gastric and jejunal feeding was not a significant difference in SAP. [35][36][37] A metaanalysis also showed that no significant difference was detected in digestive complications. [38] Meanwhile, in our study, the EN formulations were different. Six RCTs were (semi) elemental formulation, three were polymeric formulation and one was unclear. But, a meta-analysis indicated that the risk of infectious complications and mortality did not significantly differ between (semi) elemental and polymeric formulation in acute pancreatitis. [39] Third, the predicted SAP in included RCTs was defined as APACHE II score ≥ 8 or CRP levels ≥ 150 mg/L, except two RCTs defined as APACHE II score ≥ 6. [24,28] In these two study, the mean of APACHE II score was in excess of 8, so this bias may not significantly influence the definition of predicted SAP. Fourth, the intervention of the control group was not consistent in all RCTs. The intervention of four RCTs was late EN in the control group and the others were PN. But, within 48 hours after admission, our study promised the "gut rest" in the control group, and the "gut rousing" in the experimental group. [40] Nevertheless, within the constraints in some included trials, this meta-analysis suggested that EN within 48 hours after admission may be beneficial to clinical outcomes in patients with SAP or predicted SAP.

Conclusions
Early EN (starting within 48 hours after admission) significantly decreased the risk of MOF, operative intervention, systemic infections, and local septic complications based on a moderate quality of evidence, mortality with a low quality of evidence and gastrointestinal symptoms with the very low quality of evidence. In addition, early EN did not reduce the risk of SIRS based on a low quality of evidence and the other local complications with very low quality of evidence. Further, well-designed RCTs are required to explore this topic.