Study Characteristics
A total of 8092 studies were identified with 2546 duplicates (Fig. 1). Of these 133 were included for full text review yielding 14 publications which met the inclusion criteria (Table 1). A total of 408 and 414 patients were included for the continuous, or intermittent/bolus regimens respectively. Patients were typically admitted to a mixed or trauma/neurology ICU. Studies generally excluded patients with prior gastrointestinal complaints or with peritonitis. Only one study was not a randomised controlled study. Caloric estimates were generally based on a 25 to 30 kcal/kg/day requirement. Nutritional requirement outcomes were reported in only 4 studies, ranging from 23 to 82% of those included who met the prescribed intake and was consistent between groups. Similarly, illness severity scores were only reported in 7 studies, with average APACHE II scores ranging from 13 to 22.
The risk of bias in included studies varied with most studies having a moderate risk of bias, predominantly due to an absence of blinding and allocation concealment.
Outcomes
Overall, there was only a difference between continuous and intermittent/bolus administration in constipation rates, with no difference in other outcomes. Mortality was described in four studies of a total of 369 study participants (Fig. 2). No statistically significant difference was identified between intermittent/bolus and continuous EN.
There was no statistically significant difference in the number of patients colonised with potentially pathogenic bacteria in either the oropharynx or upper gastrointestinal tract, although only 3 studies of a total of 113 participants were included (Fig. 3).
Six studies of 407 participants examined pneumonia as an outcome (Fig. 4). No statistically significant difference was identified between administration methods (Fig. 4). Sensitivity analysis by removing the Bonten et al. 1996 study that defined an intermittent infusion as that administered over 18 h did not change the outcome (OR 1.25, 95% CI 0.31–5.08, p = 0.75). There was considerable heterogeneity in outcome that may be due to the variable definitions of pneumonia (Fig. 4). Pneumonia was variably defined, but the presence of blue dye in respiratory secretions was the most common method of detection.
There was no statistically significant difference between administration methods for gastrointestinal disturbance including diarrhoea (Fig. 5), constipation (Fig. 6) or increased gastric residuals (Fig. 7). Diarrhoea was assessed in 8 studies with a total of 478 study participants. No statistically significant difference was identified between continuous and intermittent/bolus EN routes. Removing the study conducted by de Arajuo et al. 2014 that defined intermittent administration as 18 h/day did not change the outcome significantly (OR 0.46, 95% CI 0.20–1.05).
In contrast to diarrhoea, constipation was only assessed in 3 studies consisting of 111 participants. There was a statistically significant difference between continuous and intermittent/bolus EN, with an increased relative risk of constipation in patients receiving continuous EN (relative risk = 2.24, 95% CI 1.01–4.97, p = 0.05) (Fig. 6).
Gastric residuals were assessed as an outcome in 5 studies (n = 223). No statistically significant difference was observed between intermittent/bolus and continuous EN (Fig. 7). Gastric residual volumes > 150 to 300 mL assessed every 3 to 4 hours were considered excessive across included studies.
Other outcomes such as glycaemic variability were assessed in two studies, but did not have standardised outcomes precluding meta-analysis. McNelly et al. assessed the incidence of hypo- and hyperglycaemia. No patients in either arm became hypoglycaemic. In contrast, 50% and 33.3% of patients in the intermittent and continuous arms became hyperglycaemic (blood glucose concentration > 10.1 mmol/L) respectively. Shahriari et al. compared the average blood glucose concentration between groups, finding no statistically significant difference (131.31 vs. 140.26 mg/dL for continuous and intermittent EN groups respectively). Three studies compared gastric pH. Overall, there was no appreciable difference between intermittent/bolus and continuous EN administration (Table 2).
Bias Assessment
There was no appreciable bias as assessed by funnel plots. The Funnel plot assessing diarrhoea is depicted in Fig. 8 as a representative sample.
Table 2
Author, Year
|
Gastric pH Continuous
|
Gastric pH Intermittent
|
Bonten, 1996
|
2.2 (IQR 1.3–3.9)
|
3.5 (IQR 1.8–5.2)
|
Gowardman, 2003
|
5
|
4
|
Spilker, 1996
|
4.7 (SD 0.5)
|
3.8 (SD 0.6)
|