Early Versus Delayed Feeding in Patients With Acute Pancreatitis

Acute pancreatitis is a common cause of hospitalization worldwide (1), and recent data suggest that incidence is increasing (2). Treatment of pancreatitis usually consists of early recognition, aggressive intravenous hydration, pain control, and supp... show more

Acute pancreatitis is a common cause of hospitalization worldwide (1), and recent data suggest that incidence is increasing (2). Treatment of pancreatitis usually consists of early recognition, aggressive intravenous hydration, pain control, and supportive monitoring. Oral intake during an episode of acute pancreatitis is believed to stimulate pancreatic exocrine activity and potentially prolong recovery. For this reason, bowel rest, including nothing-by-mouth status (NPO), has been the standard of care since at least 1950 (3). Keeping patients NPO, however, necessitates prolonged hospitalization or other forms of nutrition, both of which may increase risk for adverse events. Data suggest that enteral nutrition helps stimulate the gut and maintain its protective barrier, thus reducing bacterial overgrowth and preventing bacterial translocation and sepsis (46). Several reviews of patients with pancreatitis suggest that those who receive enteral feeding have fewer infections, shorter hospital stays, and lower mortality than those who receive total parenteral nutrition (711). In addition, enteral feeding in patients with acute pancreatitis may reduce length of hospital stay when given early (12). Despite these data, whether to feed patients with pancreatitis during an active phase of inflammation or to delay feeding until the acute phase has subsided remains controversial. Recent trials comparing early versus delayed feeding have yielded mixed results, and available guidelines are vague and contradictory (13, 14). Physician practice varies, with up to 94% of patients being placed NPO at admission (15, 16). The aim of this systematic review of randomized clinical trials was to evaluate whether early feeding affects length of hospital stay, symptoms, and clinical outcomes compared with delayed feeding in patients hospitalized with acute pancreatitis. Methods We developed a protocol (PROSPERO: CRD42015016193) and followed the PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) recommendations (17). Data Sources and Searches Between 25 March 2015 and 6 January 2017, a medical librarian (M.L.C.) performed serial literature searches for randomized controlled trials in the following databases: MEDLINE via Ovid (start date, 1950), EMBASE (start date, 1946), the Cochrane Library (start date, 1948), CINAHL (start date, 1960), and Web of Science (start date, 1996). Searches designed for each database included controlled vocabulary terms (for example, Medical Subject Headings) when available, combined with keywords to represent concepts, including pancreatitis, fast, NPO, delayed feeding, and enteral nutrition (see the Appendix for all search details). No restrictions were placed on publication date, language, or completion status. We also searched ClinicalTrials.gov (23 December 2016), hand-searched bibliographies of trials and reviews, and contacted content experts. Study Selection Randomized controlled trials were eligible for inclusion if they compared early versus delayed enteral nutrition in adults hospitalized with a diagnosis of acute pancreatitis. In keeping with prior reviews and guidelines (18, 19), we defined early feeding as feeding initiated at or within 48 hours of hospitalization and delayed feeding as feeding initiated more than 48 hours after hospitalization. Studies were excluded if they compared enteral versus parenteral nutrition, compared one form of enteral nutrition with another without varying timing of feeding, or compared one type of feeding formula with another. We considered studies published in any language and in full-text, abstract, or poster form eligible for inclusion. Two authors (V.M.V. and D.S.) independently determined study eligibility; when necessary, differences in opinion were resolved by a third author (V.C.). Interrater agreement for study eligibility and data abstraction was assessed using the Cohen coefficient. Study authors were contacted to ascertain study eligibility or clarify data. Data Extraction and Quality Assessment Data were extracted from included studies independently and in duplicate by 2 authors (V.M.V. and D.S.) using a template adapted from the Cochrane Collaboration (20). Information on the number and type of patients, definitions of early and delayed feeding, type of feeding, and outcomes (length of hospital stay, mortality, readmission, and symptoms) was extracted. Two authors (V.M.V. and D.S.) independently assessed risk of bias in included trials for the outcome of length of stay by using the Cochrane Collaboration tool (20). Studies that met low-risk criteria in each of the 6 domains were designated as having low risk of bias. Studies were classified as having unclear risk of bias if methodological data were missing. Because blinding of patients and clinicians was not feasible and blinding of data collectors and outcome adjudicators was unlikely to affect length of stay, we considered all studies to be at low risk of bias for the blinding domain. Data Synthesis and Analysis We qualitatively synthesized results of studies that included patients with mild to moderate pancreatitis separately from those that included patients with severe pancreatitis. Our main outcomes of interest were length of hospital stay, mortality, and readmission. Secondary outcomes included feeding intolerance, nausea, vomiting, recurrent abdominal pain, and necrotizing pancreatitis. We reported length of stay using the summary statistic (mean or median) provided by individual studies. One study (21) had 2 early-feeding groups, so the means and SDs of the groups were combined (20). For this study, the number of patients in each group was not provided, so an equal split was assumed. For the remaining outcomes, risk differences were calculated as the difference in the proportion of patients in each group who developed an adverse event. Because of clinical heterogeneity of patient populations, feeding protocols, and reported outcomes, meta-anal show less