The relation between nutrition and nosocomial pneumonia: randomized trials in critically ill patients.

OBJECTIVE: To review the effect of enteral nutrition on nosocomial pneumonia in critically ill patients as summarized in randomized clinical trials. STUDY IDENTIFICATION AND SELECTION: Studies were identified through MEDLINE, SCISEARCH, EMBASE, the Cochrane Library, bibliographies of primary and review articles, and personal files. Through duplicate independent review, we selected randomized trials evaluating approaches to nutrition and their relation to nosocomial pneumonia. DATA ABSTRACTION: In duplicate, independently, we abstracted key data on the design features, population, intervention and outcomes of the studies.
RESULTS: We identified four trials of enteral vs total parenteral nutrition, one trial of early enteral nutrition vs delayed enteral nutrition, one trial of gastric vs jejunal tube feeding, one trial of intermittent vs continuous enteral feeding, and three trials evaluating different enteral feeding formulae. Sample sizes were small, pneumonia definitions were variable and blinded outcome assessment was infrequent. Randomized trial evidence is insufficient to draw conclusions about the relation between enteral nutrition and nosocomial pneumonia.
CONCLUSIONS: Nutritional interventions in critically ill patients appear to have a modest and inconsistent effect on nosocomial pneumonia. This body of evidence neither supports nor refutes the gastropulmonary route of infection.

Introduction

Nosocomial pneumonia is an important cause of morbidity and mortality in hospitalized patients. Diagnosis and treatment continues to challenge clinicians and stimulate investigators. Prevention of this serious infection has been the focus of numerous studies, conferences and professional documents. Nosocomial pneumonia prevention strategies may be directed at the ventilator circuit (frequency of tubing circuit changes and gas humidification strategies), the endotracheal tube (intubation orifice, secretion drainage and suctioning) or body position (kinetic bed therapy and semirecumbancy). Other nonpulmonary approaches are pharmacologic (selective digestive decontamination and stress ulcer prophylaxis) or nutritional (the type, site and timing of enteral feeds).

The largest number of published randomized trials in intensive care medicine have evaluated selective digestive decontamination and stress ulcer prophylaxis. Five meta-analyses [1,2,3,4,5] suggest that selective digestive decontamination confers a large, clinically important and statistically significant reduction in nosocomial pneumonia rates (common odds ratio approximately 0.30, 95% CI 0.28–0.48). Nevertheless, selective digestive decontamination is not widely used, in part due to concern about long-term microbial resistance patterns and antibiotic costs [6]. Stress ulcer prophylaxis trials have been recently summarized in a meta-analysis suggesting that sucralfate, as compared with histamine-2-receptor antagonists or antacids, is associated with a trend toward a lower rate of nosocomial pneumonia (common odds ratio 0.78, 95% CI 0.60–1.01) [7]. Other experiments show that modifying gastric pH with acidified enteral feeds decreases gastric colonization, thereby supporting this underlying biologic rationale [8]. However, sucralfate is not considered of proven benefit due to the possibility that sucralfate confers a protective effect only when compared with gastric pH-altering drugs (which themselves are associated with a modest increase in nosocomial pneumonia compared to control) [7].

Kinetic bed therapy has been reviewed in a meta-analysis of six trials in seriously and critically ill patients, which indicated a significantly lower rate of pneumonia and atelectasis in patients receiving continuous postural oscillation [9]. A less expensive and adaptable pneumonia prevention strategy focussing on body position has been studied in three randomized trials [10,11,12]. Torres et al [10] found that after instillation of radioactive technetium sulfur colloid into the stomach, radioactive counts in endobronchial secretions were significantly higher in samples obtained while patients were supine than when they were semirecumbent. In another study, scintigraphic evidence of esophageal reflux was found in 81% of patients in the supine position compared to 64% in the semirecumbent position [11]. Orozco-Levi et al administered nasogastric technetium sulfur colloid and found that radioactive counts in endobronchial secretions increased over time, but were higher in the supine than the semirecumbent position [12]. Although a causal relationship between pneumonia and this secondary endpoint of aspiration of gastric contents has not been convincingly demonstrated, these trials are in keeping with the gastropulmonary route of infection.

The gastropulmonary route of infection is a concept at least two decades old [13], support for which is derived from multiple human observational studies and experimental evidence [14, 15]. Enteral nutrition, compared to parenteral nutrition, is associated with decreased translocation in animals and decreased infectious morbidity in critical illness in humans [16]. Accordingly, it holds the promise of affording protection against nosocomial pneumonia. However, enterally feeding critically ill patients is often associated with intolerance, thereby predisposing them to aspiration pneumonia. The goal of this systematic review is to critically appraise and summarize the randomized trials of nutritional strategies and their influence on nosocomial pneumonia in critically ill patients.

Methods

Study identification

To identify randomized trials, we searched two computerized databases from 1980 onwards. For MEDLINE, we used the following text words and keywords: critical care, intensive care units, pneumonia, respiratory tract infection, mechanical ventilation, gastropulmonary, enteral nutrition, randomized controlled trials, prospective studies. For EMBASE, we used: pneumonia, prevention, control. Frequently cited articles were identified and SCISEARCH (Science Citation Index online) was used to locate any additional relevant randomized trials. We also used the Cochrane Library, searching the Clinical Trials Registry for randomized trials, and the Cochrane Database of Systematic Reviews (CDSR) as well as the Database of Abstracts of Reviews (DARE) for systematic reviews containing relevant primary studies. We confined our search to studies enrolling non-neutropenic adult humans without the human immunodeficiency virus. We had no language restrictions.

The titles (and the abstracts, when available) in the MEDLINE and EMBASE printouts, and the reference lists of all primary and review articles were reviewed independently in duplicate. Any additional relevant articles were thereby identified and retrieved.

Study selection

The following selection criteria were applied to the full manuscripts by two reviewers independently:

1. Population: critically ill adults, including trauma and burn patients.

2. Interventions: nutritional support.

3. Outcomes: nosocomial pneumonia.

4. Design: published randomized trials in humans.

A priori, we excluded relevant nutritional interventions in seriously but not necessarily critically ill patients, studies examining surrogate endpoints for pneumonia [8], studies which did not report how pneumonia was diagnosed [17,18,19], studies which evaluated or reported composite infectious outcomes [20], and duplicate publications [21].

Study characteristics and data abstraction

Two reviewers abstracted data from the randomized trials to describe the method of treatment allocation, the proportion of patients who were excluded post-randomization, whether cointerventions were described, whether the endpoints were assessed by investigators blinded to the intervention, and the outcome definitions employed. Disagreements between reviewers on design characteristics and raw data abstraction were resolved by discussion and consensus.

Analysis

We measured agreement between reviewers on the selection of articles for inclusion in the review. We standardized presentation of the randomized trial results using relative risk, and calculated 95% confidence intervals using the log transformation method. Since study questions and trial designs differed, we did not statistically pool results of these trials, or subgroups of them, in a meta-analysis.

Results

Study identification and selection

The search yielded four trials of enteral vs total parenteral nutrition [22,23,24,25], one trial of early enteral nutrition vs delayed enteral nutrition [26], one trial of gastric vs jejunal tube feeding [27], one trial of intermittent vs continuous enteral feeding [28], and three trials evaluating different enteral feeding formulae [29,30,31]. Agreement was 100% for selection of these trials and systematic reviews.

Study characteristics

Study characteristics are reported in Table 1. Patients were medical or surgical ICU patients, burn, or trauma victims. Two studies were explicit about concealment of randomization using sealed envelopes [26,28]. The nature of some of these comparisons precluded blinding of patients and caregivers. Patients were unlikely to be aware of the details of their care and were not participating in assessment of the presence of nosocomial pneumonia. However, lack of blinding could have affected the care delivered by bedside nurses, respiratory therapists and intensivists, which could have affected the development of lung infection. In one trial, the neurosurgeon evaluating outcomes was blinded [22]; in another, confirmation of outcome was conducted by a second blinded surgeon [24]. Two of the three studies comparing different feeding products employed blinded outcome assessment [29,31].

Table 1

Nutrition and nosocomial pneumonia: study characteristics

Author [reference]	Intervention	Population	Allocation	Cointerventions	Exclusion post-	Blinding of	Definition of
					randomization	outcome accessor	VAP
Young et al [22]	Nasogastric	Head injury	`Was randomly	All patients received	7 Exclusions:	Neurosurgen	Infiltrate and
	enteral nutrition	patients	assigned to'	prokinetic	5-early death, 2	evaluating	leukocytosis
	vs total				-withdrew	outcomes was	premature cells,
	parenteral					blinded	fever, positive
	nutrition						sputum culture
Moore et al [23]	Enteral nutrition	Trauma patients	`Randomized by	Broad spectrum	No exclusions:	Outcome	New infiltrate and
	via needle	requiring	computer	antibiotics to both	4-early death,	assessment not	fever, leukocytosis
	catheter	emergency	assignment'	groups	3-reoperation,	blinded	and purulent
	jejunostomy vs	celiotomy			3-chronic illness,		sputum
	total parenteral				2-ATI> 40,
	nutrition				2-head injury,
					1-mechanical
					failure,
					1-transfer
Kudsk et al [24]	Enteral nutrition	Patients with blunt	`Computer	NR	2 Exclusions:	Secondary	New infiltrate and
	via needle	and penetrating	generated		death within 4	confirmation of	leukocytosis,
	jejunostomy vs	abdominal trauma	randomization		days	outcome by	positive sputum or
	total parenteral		table'			blinded surgeon	BAL, or purulent
	nutrition						sputum
Borzotta et al [25]	Enteral nutrition	Patients with	`Computer	Jejunostomy group	NR	Outcome	Infiltrate and
	via needle	severe closed head	generated random	had gastrostomy		assessment not	fever, leukocytosis,
	catheter	injury	number	tube drainage		blinded	leukorrhea and
	jejunostomy vs		assignment'				bacteria on Gram
	total parenteral						stain
	nutrition
Eyer et al [26]	Early (<24 h)	Patients with blunt	`Randomization by	All patients received	14 Exclusions:	Outcome	New infiltrate and
	nasoduodenal	abdominal trauma	card drawn from	either sucralfate or	3-regular diet,	assessment not	significant growth
	tube feeding vs		sealed envelope'	antacids but group not	3-steroids,	blinded	on sputum
	late (>72 h)			specified	2-no NGT,		culture with <10
	nasoduodenal				6-miscellaneous		epithelial cells,
	tube feeding						>25 wbc/hpf OR
							purulent
							secretions, fever
							and leukocytosis
Montecalvo et al [27]	Gastric vs jejunal	Medical and	`Randomly	25 Patients received	5 Patients crossed	Cultures	New and
	tube feeding	surgical ICU	assigned	sucralfate; 1 H2RA;	over from jejunal to	reviewed blinded	persistent
		patients	according to	2 H2RA and antacids;	gastric group and	to group	infiltrate and
			computer	8 sucralfate and	2 patients crossed	assignment	three of: purulent
			generated random	either H2RA or	over from gastric		sputum with
			number code'	antacids; 1 no stress	to jejunal group;		numerous
				ulcer prophylaxis, but	these 7 patients		bacteria, purulent
				group not specified	were included until		sputum with
					the day they		nosocomial
					crossed over		pathogen, T>386,
							or wbc >10
Bonten et al [28]	Intermittent	Mixed ICU	`Randomization	Intermittent: 13-	None	Outcome	New and
	enteral feeding	patients and	was performed	antacids and 17-		assessment not	persistent
	(18 h) vs	cardiac surgery	with sealed	sucralfate;		blinded	infiltrate and 3 of:
	continuous	patients needing	envelopes'	continuous: 7 -			T>38 or T<355
	enteral feeding	ventilation > 3		antacids and 23 -			OR wbc > 10
	(24 h)	days		sucralfate			and/or left shift
							or wbc < 3 OR 10
							wbc/hpf on ET
							Gram strain OR
							positive ET
							aspirate and one
							of these: BAL
							(positive if > 104
							CFU/ml) OR
							PSB (positive if
							>103 CFU/ml)
							OR positive
							blood culture OR
							positive pleural
							culture
Gottsschlich et al [29]	Modular tube	Burn patients	`Random number	NR	NR	Physicians,	Infiltrate and
	feeding vs two	(>10% BSA)	table stratified for			nurses,	positive sputum
	standard enteral		age, center and			technicians,	culture and
	feeding		burn size'			clinical and	systemic
	(Osmolite vs					research	antibotics
	Traumacal)					personnel were
						blinded
Moore et al [30]	Early enteral	Trauma patients	`Randomized by a	NR	16 exclusions:	Outcome	New and
	immune-		computer-		9-inappropriate	assessment not	progressive
	enhancing		generated		randomizations,	blinded	infiltrate, fever,
	feeding vs		schedule'		7-drop -outs		leukocytosis,
	standard enteral				1-early death		positive sputum
	feeding						Gram stain with
	(Vivonex)						many polys
Kudsk et al [31]	Early immune-	Trauma patients	`Computer-	Short-term broad	NR	All caregivers	New or changing
	enhancing	requiring	generated	spectrum antibiotics		blinded except	infiltrate and
	feeding via	emergency	randomization	to both groups		nutritionist	fever,
	jejunostomy vs	celiotomy	table'				leukocytosis,
	standard enteral						purulent sputum
	feeding						underwent BAL
	(Promote)						(positive if > 103
							CFU/hpf)

Abbreviations: ATI=acute trauma index; BAL=bronchoalveolar lavage; NGT=nasogastric tube; wbc=white blood cells; hpf=high power field; H2RA=histamine-2-receptor antagonists; ET=endotracheal; CFU=colony forming units; BSA=body surface area; NR=not reported; VAP=ventilator-associated pneumonia; PSB=protected specimen brush.

Cointerventions are interventions which are unrelated to the study question, yet may impact on the outcome, and could be unequally distributed across groups. These include stress ulcer prophylaxis and selective digestive decontamination (Table 1). Other cointerventions not mentioned, but potentially important to standardize or report, might include chest physiotherapy and position of the patients.

In two trials, the pneumonia definition incorporated but did not require invasive bronchoscopic techniques [24,28]; in a third trial, a positive bronchoalveolar lavage was required for the diagnosis [31].

Study results

The results of these randomized trials are presented in Table 2. The four trials evaluating total parenteral vs enteral nutrition yield inconsistent results. In one, there was a trend toward a lower rate of pneumonia associated with enteral nutrition [23], in another study the pneumonia rate was significantly lower in the enteral nutrition group [24], and in the remaining two studies, the pneumonia rate was slightly higher in patients receiving enteral nutrition [22,25].

Table 2

Results of randomized trails of nutrition and nosocomial pneumonia

Intervention (author [reference])	Pneumonia rates	Relative risk (95% Cl)
Nasogastric enteral nutrition vs parenteral nutrition (Young [22])	TPN: 6/23 (26%)	1.23 (0.51–2.95)
	EN: 9/28 (32%)
Jejunostomy feeding vs total parenteral nutrition (Moore [23])	TPN: 6/30 (20%)	Undefined
	EN: 0/29 (0%)
Jejunostomy feeding vs total parenteral nutrition (Kudsk [24])	TPN: 14/45 (31%)	0.38 (0.16–0.90)
	EN: 6/51 (12%)
Jejunostomy feeding vs total parenteral nutrition (Borzotta [25])	TPN: 9/23 (39%)	1.06 (0.56–2.02)
	EN: 15/36 (42%)
Early nasoduodenal vs late nasoduodenal feeding (Eyer [26])	Late: 4/19 (21%)	2.00 (0.72–5.54)
	Early: 8/19 (42%)
Jejunal vs gastric feeding (Montecalvo [27])	Gastric: 2/19 (11%)	Undefined
	Jejunal: 0/19 (0%)
Intermittent enteral feeding vs continuous enteral feeding (Bonten [28])	CEF: 5/30 (17%)	1.0 (0.32–3.10)
	IEF: 5/30 (17%)
Modular tube feeding (MTF) vs Osmolite vs Traumacal (Gottschlich [29])	Osmolite: 6/14 (43%)	0.27 (0.07–1.15)*
	Traumacal: 9/19 (47%)	0.25 (0.06-0.99)†
	MTF: 2/17 (12%)
Immun-Aid vs Vivonex (Moore [30])	Vivonex: 4/47 (9%)	0.92 (0.24–3.48)
	Immun-Aid: 4/51 (8%)
Immun-Aid vs Promote (Kudsk [31])	Promote: 3/17 (18%)	Undefined
	Immun-Aid: 0/16 (0%)

Abbreviations: EN = enteral nutrition; TPN = total parenteral nutrition; CEF = continuous enteral feeding; IEF = intermittent enteral feeding. * Osmolite compared to MTF. † Traumacal compared to MTF.

One study examined early enteral nasoduodenal nutrition begun within 24 h vs nasoduodenal enteral nutrition delayed for 72 h. In patients receiving early feeds, there was a trend toward increased pneumonia (8/19 vs 4/19, respectively) [26].

Considering the potential for enteral nutrition to cause aspiration pneumonia, one study tested the effect of proximal vs distal delivery sites [27]. Two cases of pneumonia were identified amongst those 19 patients receiving prepyloric gastric feeds and no cases were observed in the 19 patients receiving post-pyloric feeds through a jejunal tube.

To avoid continuous alkalinization and intragastric Gram-negative growth associated with enteral feeding, intermittent enteral nutrition was compared with continuous enteral nutrition in one trial [28]. Five of 30 patients in each group developed nosocomial pneumonia.

Three studies examined different enteral feeding formulae and their relation to lung infection. The first compared modular tube feeds (a high protein, low fat, linoleic acid-restricted formulation enhanced with arginine, cysteine, vitamin A, zinc, omega-3-polyunsaturated fatty acids, and ascorbic acid) against Osmolite and Traumacal and found a trend toward lower pneumonia rates in the modular tube feed patients [29]. There was no difference in pneumonia between trauma patients fed Immun-Aid (containing glutamine, arginine, omega-3-polyunsaturated fatty acids, nucleotides, and branched chain amino-acids) vs Vivonex (standard enteral formulae) [30]. In another study of trauma patients, Immun-Aid was associated with a trend toward a lower pneumonia rate than patients fed with Promote (an isonitrogenous, isocaloric diet) [31].

Discussion

The results of these 10 trials of feeding strategies, either individually or in aggregate, provide inconclusive evidence about the relation between enteral nutrition and nosocomial pneumonia. These studies enrolled a total of 582 patients and contribute 117 cases of pneumonia. The single trial showing a significantly lower pneumonia rate associated with jejunal enteral nutrition over parenteral nutrition [24] has not been translated into widespread clinical policy, perhaps due to the inconvenience and expertise required for jejunostomy tubes. Aside from concerns about type I and II error when interpreting the trials in this review, there are other relevant outcomes addressed by some, but not all of these studies, including effects on nutritional markers and adverse outcomes such as catheter sepsis and patient comfort. Readers are referred to the original articles for these important details.

Factors such as cost, and ease with which the feeding strategy can be employed, are additional issues that bear on the interpretation and application of these trial results in practice. Intensivists also consider evidence from observational studies when making clinical decisions. Given these provisos, it is not surprising that definitive statements about enteral nutrition and lung infection are not forthcoming. Some guidelines from the Center for Disease Control on the prevention of nosocomial pneumonia [32] focus on gastropulmonary approaches. Stress ulcer prophylaxis with an agent that does not increase gastric pH was `suggested for implementation in many hospitals and supported by suggestive clinical and epidemiologic studies and a strong theoretical rationale'. Other interventions labelled as `unresolved for which no recommendations were made' included jejunal feeding, intermittent enteral feeding and selective digestive decontamination. In the American Thoracic Society statement on prevention of hospital-acquired pneumonia in adults [33], some prophylactic interventions were classified as having `probable effectiveness, used widely in some clinical settings', such as distal enteral nutrition, semi-erect positioning, and sucralfate. Selective digestive decontamination was considered `of unproven value used on a limited investigational or clinical basis'.

Nutrition is integral to the care of an ICU patient. The method, site and timing of enteral nutrition may have a protective or predisposing influence on the risk of nosocomial infection [34], though strong proof from experiments in humans does not currently exist. Although a meta-analysis of published and unpublished trials of general surgical and trauma patients suggested a lower pneumonia rate in patients receiving enteral nutrition vs total parenteral nutrition [35], published data from ventilated medical ICU patients are sparse, and generalizing to other populations may not be reasonable. Interventions requiring further investigation with large rigorous studies of ICU patients include those discussed in this review, as well as the size of feeding tubes [36], their insertion site, where the tubes are located in the gastrointestinal tract [37], feeding advancement schedules, and the effect of prokinetic drugs [38].