Results of the search

We identified a total of 1999 citations. Through reading the titles and then abstracts, we obtained the full texts of 89 citations that were potentially eligible for inclusion in our review. Of those 89 citations, 25 did not meet our inclusion criteria and were excluded for the reasons described in the table Characteristics of excluded studies. Six ongoing clinical trials were identified through searching the websites clinicaltrials.gov and controlled‐trials.com (Characteristics of ongoing studies). One study (Hallay 2002) is waiting assessment as we were unable to obtain the full text. Finally, we included 57 articles from 53 studies (four articles were secondary publications) in this review (Figure 1).

Figure 1

---

---

Study flow diagram.

Included studies

These 53 studies enrolled a total of 4671 patients. All participants were adults with a critical illness or undergoing elective major surgery; they were allocated to either a glutamine supplementation group or control group. All studies reported at least one of the outcomes which we had predefined in our protocol. The studies fell broadly into the following groups.

Excluded studies

We excluded 25 clinical studies for the reasons described in the Characteristics of excluded studies.

Allocation

Eighteen studies clearly described the process for adequate generation of a randomized sequence and concealment of allocation and were classified as low risk of bias. Random sequence generation or allocation concealment was rated as having an unclear risk of bias in the remaining 35 studies due to not reporting sequence generation or allocation details. No study was rated as high risk of bias (Figure 3).

Blinding

Incomplete outcome data

In nine studies the number of patients withdrawn after randomization was high (4 to 49) and the studies were judged as having a high risk for attrition bias (Conejero 2002; Estívariz 2008; Garrel 2003; Goeters 2002; Hajdú 2012; Kłek 2005; Luo 2008; Mertes 2000; Yang 2008). Two studies were rated at unclear risk: one study planned to exclude the patients who died within 72 hours of admission but did not report the number of patients excluded (Pattanshetti 2009), another study withdrew five patients without giving reasons (Wernerman 2011). For the remaining 42 studies the risk of attrition bias was low (Figure 2; Figure 3).

Selective reporting

We judged two studies as high risk for reporting bias. One study did not report the length of hospital stay outcome completely (Qiu 2009), the other study did not report the length of ICU stay and all‐cause six‐month mortality outcomes completely (Wernerman 2011). We were, therefore, unable to enter these two studies in a meta‐analysis. Three studies were judged as having an unclear risk for reporting bias. One study only reported the average length of stay and mortality outcomes in the subgroup of patients treated for longer than nine days, and whether this was pre‐specified was unclear (Goeters 2002). Two other studies either did not report baseline infection associated blood test results (Pattanshetti 2009) or did not report organ function tests (Zhu 2000) and were therefore judged at unclear risk. The remaining 48 studies were free from selective reporting bias (Figure 2; Figure 3).

Other potential sources of bias

One study had baseline bias and was rated at high risk. This was because the patients in the treatment group were younger than those in the control group (Kumar 2007). In another study, patients that were clinically accepted for parenteral nutrition were included in the trial, it was unclear whether the patients were all critically ill. This study was rated at unclear risk (Powell‐Tuck 1999). No other significant potential sources of bias were identified in the remaining trials (Figure 2; Figure 3).

Infectious complications

Thirty‐three trials (including 2303 patients) provided data on the nosocomial infectious complications (see Characteristics of included studies). The absolute risk for infectious complications was 39.9% (458/1148) in the control group and 31.8% (367/1155) in the intervention group. Pooling of these data suggested that glutamine supplementation reduced the infectious complications rate in adults with critical illness or undergoing elective major surgery (RR 0.79, 95% CI 0.71 to 0.87, P ＜ 0.00001, I² = 8%) (Analysis 1.1). By setting the assumed control risk at 40%, we calculated the number needed to treat for an additional beneficial outcome as 12. The funnel plot for this outcome was asymmetrical, suggesting that smaller studies with outcomes favouring non‐supplemented patients appeared to be lacking. One newly published large sample, randomized blinded trial (Heyland 2013) provided data on the infectious outcome in glutamine supplementation patients together with patients on antioxidant supplementation. We were told by the author that the detailed infectious incidence in glutamine supplementation patients will be published in the future (Heyland 2013 [pers comm]). This study was not included in the analysis. We also used the trim and fill method for this outcome to identify and correct for funnel plot asymmetry arising from publication bias. The method suggested that 13 studies are missing. Using a fixed‐effect model (inverse variance method) the original point estimate (RR) and 95% CI for the combined studies was 0.84 (95% CI 0.80 to 0.92). Using trim and fill the imputed point estimate was 0.89 (95% CI 0.81 to 0.98). Using the random‐effects model the RR was 0.83 (95% CI 0.75 to 0.92). Using trim and fill the imputed point estimate was 0.87 (95% CI 0.77 to 0.99). These results suggested that the combined results were not changed after adding the missing studies.

Sensitivity analysis showed that the risk of infectious complications in the low risk of bias studies (Andrews 2011; Déchelotte 2006; Dong 2008; Grau 2011; Hall 2003; Zhou 2003; Zhou 2004) was not different between the groups (RR 0.85, 95% CI 0.72 to 1.01, P = 0.06, I² = 9%). The difference in combined effect size between the main analysis and the sensitivity analysis could be explained by the reduced sample size but more importantly the change in RR from 0.79 in the main analysis to 0.85 in the sensitivity analysis suggested that the effect size could have been partly explained by within study biases.

Infectious complications: elective surgery patients versus critically ill patients

Eleven trials (Dong 2008; Estívariz 2008; Jiang 1999; Karwowska 2001; Kłek 2005; Lu 2011; Neri 2001; O'Riordain 1994; Spittler 2001; Wang 2006; Zhu 2000) included 456 patients who were enrolled for elective surgery. The pooled RR was 0.59 (95% CI 0.41 to 0.86, P = 0.005, I² = 0%). Twenty‐two trials (Andrews 2011; Conejero 2002; Cucereanu‐Badica 2013; Déchelotte 2006; Eroglu 2009; Fuentes‐Orozco 2004; Fuentes‐Orozco 2008; Garrel 2003; Grau 2011; Hajdú 2012; Hall 2003; He 2004; Kumar 2007; Ockenga 2002; Pérez‐Bárcena 2008; Pérez‐Bárcena 2010; Powell‐Tuck 1999; Sahin 2007; Wischmeyer 2001; Zhao 2013; Zhou 2003; Zhou 2004) that enrolled 1847 critically ill patients were pooled to give a RR of 0.82 (95% CI 0.73 to 0.91, P = 0.001, I² = 18%) (Analysis 2.1).

Infectious complications: enteral glutamine supplementation versus parenteral glutamine supplementation

Twenty‐eight studies (Andrews 2011; Cucereanu‐Badica 2013; Déchelotte 2006; Dong 2008; Eroglu 2009; Estívariz 2008; Fuentes‐Orozco 2004; Fuentes‐Orozco 2008; Grau 2011; Hajdú 2012; He 2004; Jiang 1999; Karwowska 2001; Kłek 2005; Lu 2011; Neri 2001; Ockenga 2002; O'Riordain 1994; Pérez‐Bárcena 2008; Pérez‐Bárcena 2010; Powell‐Tuck 1999; Sahin 2007; Spittler 2001; Wang 2006; Wischmeyer 2001; Zhao 2013; Zhou 2004; Zhu 2000) examined the effects of parenteral glutamine supplementation on infection. Meta‐analysis showed a lower risk of infectious complications in patients receiving parenteral glutamine supplementation than those without glutamine supplementation (RR 0.78, 95% CI 0.69 to 0.87, P ＜ 0.00001, I² = 0%). Five studies (Conejero 2002; Garrel 2003; Hall 2003; Kumar 2007; Zhou 2003) tested the effects of enteral glutamine supplementation on the infection rate. There was no significant difference between the treatment and control groups (RR 0.83, 95% CI 0.67 to 1.02, P = 0.07, I² = 50%) (Analysis 3.1). The test for subgroup differences was not statistically significant (P = 0.61).

Infectious complications: patients with malignant disease versus patients with non‐malignant disease

Six studies (Kłek 2005; Lu 2011; Neri 2001; O'Riordain 1994; Wang 2006; Zhu 2000) included surgical patients with malignant disease and the pooled RR was 0.45 (95% CI 0.23 to 0.90, P = 0.02, I² = 0%). Seventeen studies (Conejero 2002; Cucereanu‐Badica 2013; Eroglu 2009; Estívariz 2008; Fuentes‐Orozco 2004; Fuentes‐Orozco 2008; Garrel 2003; Hajdú 2012; He 2004; Karwowska 2001; Ockenga 2002; Pérez‐Bárcena 2010; Sahin 2007; Wischmeyer 2001; Zhao 2013; Zhou 2003; Zhou 2004) enrolled patients with non‐malignant disease. The pooled analysis showed that the risk of infectious complications was lower in the glutamine supplemented patients than in the patients that did not receive the supplement (RR 0.66, 95% CI 0.55 to 0.78, P < 0.00001, I² = 0%). The remaining 10 studies (Andrews 2011; Déchelotte 2006; Dong 2008; Grau 2011; Hall 2003; Jiang 1999; Kumar 2007; Pérez‐Bárcena 2008; Powell‐Tuck 1999; Spittler 2001) enrolled patients with uncertain or mixed disease. The pooled RR was 0.91 (95% CI 0.80 to 1.03, P = 0.15, I² = 0%) (Analysis 4.1).

Mortality

Thirty‐seven studies (3494 patients) reported on mortality. Of these 37 studies, 36 (Andrews 2011; Çekmen 2011; Chai 2006; Conejero 2002; Cucereanu‐Badica 2013; Déchelotte 2006; Dong 2008; Eroglu 2009; Estívariz 2008; Fuentes‐Orozco 2004; Fuentes‐Orozco 2008; Garrel 2003; Goeters 2002; Grau 2011; Griffiths 1997; Hajdú 2012; Hall 2003; He 2004; Heyland 2013; Karwowska 2001; Kumar 2007; Lin 2002; Neri 2001; Ockenga 2002; Pattanshetti 2009; Pérez‐Bárcena 2008; Pérez‐Bárcena 2010; Powell‐Tuck 1999; Qiu 2009; Sahin 2007; Tian 2006; Wang 2006; Wernerman 2011; Wischmeyer 2001; Yang 2008; Zhao 2013) reported short‐term (hospital or closest to one month) mortality. The combined rate of mortality from these studies was not statistically different between the group which received a glutamine supplement and the group which did not (RR 0.89, 95% CI 0.78 to 1.02, P = 0.10, I² = 22%) (Analysis 1.2). Eleven studies (Andrews 2011; Déchelotte 2006; Goeters 2002; Grau 2011; Griffiths 1997; Hall 2003; Heyland 2013; Powell‐Tuck 1999; Qiu 2009; Tjäder 2004; Wernerman 2011) reported long‐term (closest to six months) mortality. Meta‐analysis of these 11 studies yielded an RR of 1.00 (95% CI 0.89 to 1.12, P = 0.94, I² = 30%) (Analysis 1.3). The funnel plot for the short‐term mortality outcome was asymmetrical, suggesting that smaller studies with outcomes favouring non‐supplemented patients appeared to be lacking; while the funnel plot for the long‐term mortality outcome was symmetrical. The trim and fill method suggested that 10 studies were missing for the short‐term mortality outcome, and results indicated that the combined results were not changed after filling in these missing studies. Using the fixed‐effect model (inverse variance method) the original point estimate (RR) for the combined studies was 0.93 (95% CI 0.81 to 1.07). Using the trim and fill method the imputed point estimate was 1.00 (95% CI 0.87 to 1.14). Using the random‐effects model the RR for the combined studies was 0.83 (95% CI 0.69 to 1.00). Using trim and fill the imputed RR was 0.93 (95% CI 0.76 to 1.15).

Sensitivity analysis showed that both the short‐ and long‐term mortality rates in low risk of bias studies were not different between the groups (RR 1.10, 95% CI 0.93 to 1.30, P = 0.29, I² = 21%; RR 0.99, 95% CI 0.80 to 1.23, P = 0.96, I² = 46% respectively).

Short‐term mortality: elective surgery patients versus critically ill patients

Twenty‐nine studies (Andrews 2011; Çekmen 2011; Chai 2006; Conejero 2002; Cucereanu‐Badica 2013; Déchelotte 2006; Eroglu 2009; Fuentes‐Orozco 2004; Fuentes‐Orozco 2008; Garrel 2003; Goeters 2002; Grau 2011; Griffiths 1997; Hajdú 2012; Hall 2003; He 2004; Heyland 2013; Kumar 2007; Ockenga 2002; Pattanshetti 2009; Pérez‐Bárcena 2008; Pérez‐Bárcena 2010; Powell‐Tuck 1999; Sahin 2007; Tian 2006; Wernerman 2011; Wischmeyer 2001; Yang 2008; Zhao 2013) enrolling 3157 critically ill patients reported short‐term mortality. The pooled RR was 0.91 (95% CI 0.79 to 1.04, P = 0.15, I² = 19%). The mortality outcome was reported in seven studies (Dong 2008; Estívariz 2008; Karwowska 2001; Lin 2002; Neri 2001; Qiu 2009; Wang 2006) enrolling patients for elective surgery. No deaths occurred during the study period in six of the seven studies so the data could not be combined (Analysis 2.2).

Short‐term mortality: enteral glutamine supplementation versus parenteral glutamine supplementation

Five clinical trials (Conejero 2002; Garrel 2003; Hall 2003; Kumar 2007; Pattanshetti 2009) examined the effects of enteral glutamine supplementation on short‐term mortality. Pooling of these studies suggested no difference in mortality rate between the intervention and control groups (RR 0.91, 95% CI 0.64 to 1.30, P = 0.61, I² = 48%). Thirty trials (Andrews 2011; Çekmen 2011; Chai 2006; Cucereanu‐Badica 2013; Déchelotte 2006; Dong 2008; Eroglu 2009; Estívariz 2008; Fuentes‐Orozco 2004; Fuentes‐Orozco 2008; Goeters 2002; Grau 2011; Griffiths 1997; Hajdú 2012; He 2004; Karwowska 2001; Lin 2002; Neri 2001; Ockenga 2002; Pérez‐Bárcena 2008; Pérez‐Bárcena 2010; Powell‐Tuck 1999; Qiu 2009; Sahin 2007; Tian 2006; Wang 2006; Wernerman 2011; Wischmeyer 2001; Yang 2008; Zhao 2013) used parenterally supplemented glutamine. We pooled the data on the 2207 patients enrolled in these studies, which yielded a RR of 0.75 (95% CI 0.63 to 0.90, P = 0.002, I² = 0%) (Analysis 3.2). Only one study (Heyland 2013) supplemented glutamine both parenterally and enterally but this study could not be entered into a subgroup analysis.

Short‐term mortality: patients with malignant disease versus patients with non‐malignant disease

Three studies (Lin 2002; Neri 2001; Wang 2006) included 120 surgical patients with malignant disease. No deaths occurred during the study period in these three studies, so the RR could not be estimated. Nineteen studies (Chai 2006; Conejero 2002; Cucereanu‐Badica 2013; Eroglu 2009; Estívariz 2008; Fuentes‐Orozco 2004; Fuentes‐Orozco 2008; Garrel 2003; Griffiths 1997; Hajdú 2012; He 2004; Karwowska 2001; Ockenga 2002; Pattanshetti 2009; Pérez‐Bárcena 2010; Sahin 2007; Wischmeyer 2001; Yang 2008; Zhao 2013) enrolled 921 patients with non‐malignant disease. The pooled analysis of these studies showed a lower mortality risk for participants receiving glutamine supplementation compared with those receiving placebo (RR 0.61, 95% CI 0.46 to 0.79, P = 0.002, I² = 0%). The remaining 14 studies (Andrews 2011; Çekmen 2011; Déchelotte 2006; Dong 2008; Goeters 2002; Grau 2011; Hall 2003; Heyland 2013; Kumar 2007; Pérez‐Bárcena 2008; Powell‐Tuck 1999; Qiu 2009; Tian 2006; Wernerman 2011) enrolled 2413 patients with uncertain or mixed disease, and the pooled RR was 1.02 (95% CI 0.87 to 1.20, P = 0.78, I² = 26%) (Analysis 4.2).

Long‐term mortality

Data on 124 patients with non‐malignant disease from two studies comparing nutrition with glutamine supplementation to nutrition without glutamine supplementation (Griffiths 1997; Tjäder 2004) were pooled to give a RR of 0.69 (95% CI 0.48 to 1.00, P = 0.05, I² = 0%). The remaining subgroup analysis did not find a mortality difference between the groups (Analysis 2.3; Analysis 3.3; Analysis 4.3).

Length of hospital stay

Hospital length of stay was reported in 36 studies (see Characteristics of included studies). Of these 36 studies, 25 (Cucereanu‐Badica 2013; Dong 2008; Estívariz 2008; Fuentes‐Orozco 2004; Fuentes‐Orozco 2008; Garrel 2003; Goeters 2002; He 2004; Jiang 1999; Karwowska 2001; Mertes 2000; Morlion 1998; Neri 2001; Pattanshetti 2009; Peng 2004; Pérez‐Bárcena 2008; Sahin 2007; Spittler 2001; Wang 2006; Yang 2008; Yao 2007; Zhao 2013; Zhou 2002; Zhou 2003; Zhou 2004) reported the outcome as mean (standard deviation or standard error of mean). Pooled analysis of these studies showed that the length of hospital stay was shorter in the glutamine supplementation group than in the control group (mean difference (MD) ‐4.14, 95% CI ‐5.46 to ‐2.81, P < 0.0001, I² = 70%). In the remaining 11 studies (Andrews 2011; Déchelotte 2006; Grau 2011; Hajdú 2012; Hall 2003; Heyland 2013; Kumar 2007; Kłek 2005; Ockenga 2002; Pérez‐Bárcena 2010; Powell‐Tuck 1999) the data were reported as the median (interquartile range) or median (range). The median (range) in each study was converted into mean (standard deviation) by using the formula from Hozo 2005, while the estimated standard deviation was calculated from the interquartile range based on the assumption that the width of the interquartile range was approximately 1.35 standard deviations (Handbook 2011). All 36 studies were included in the meta‐analysis after conversion and the results suggested that the length of hospital stay was shorter in the intervention group than in the control group (MD ‐3.46, 95% CI ‐4.61 to ‐2.32, P < 0.0001, I² = 63%) (Analysis 1.4). The funnel plot for this outcome was symmetrical.

Sensitivity analysis, including eight low risk of bias studies (Andrews 2011; Déchelotte 2006; Dong 2008; Grau 2011; Hall 2003; Heyland 2013; Zhou 2003; Zhou 2004), also showed that glutamine supplementation had beneficial effects in reducing the length of hospital stay (MD ‐2.9, 95% CI ‐5.3 to ‐0.5, P = 0.02, I² = 58%).

The substantial heterogeneity between studies for this outcome may have been because these studies included participants with different categories of disease. Hence we did the following subgroup analysis according to types of diseases. Among the 36 trials, seven trials included burns patients (Cucereanu‐Badica 2013; Garrel 2003; Pattanshetti 2009; Peng 2004; Zhou 2002; Zhou 2003; Zhou 2004) and seven trials included acute pancreatitis patients (Fuentes‐Orozco 2008; Hajdú 2012; He 2004; Ockenga 2002; Sahin 2007; Yang 2008; Zhao 2013). Meta‐analysis of these two subgroups suggested beneficial effects of glutamine in reducing the length of hospital stay, with mild to moderate heterogeneity between studies (MD ‐7.24, 95% CI ‐11.28 to ‐3.21, P = 0.0004, I² = 49% for burn patients; MD ‐1.75, 95% CI ‐2.42 to ‐1.08, P < 0.00001, I² = 0% for acute pancreatitis patients). However, combining the data from nine trials that included critically ill patients other than burns and acute pancreatitis patients (Andrews 2011; Déchelotte 2006; Goeters 2002; Grau 2011; Hall 2003; Heyland 2013; Pérez‐Bárcena 2008; Pérez‐Bárcena 2010; Powell‐Tuck 1999) revealed that there was no difference between the treatment and control groups, with moderate heterogeneity between studies (MD ‐1.16, 95% CI ‐4.05 to 1.72, P = 0.43, I² = 25%). A meta‐analysis of 13 trials including surgical patients (Dong 2008; Estívariz 2008; Fuentes‐Orozco 2004; Jiang 1999; Karwowska 2001; Kumar 2007; Kłek 2005; Mertes 2000; Morlion 1998; Neri 2001; Spittler 2001; Wang 2006; Yao 2007) showed that there was still substantial heterogeneity among these studies (MD ‐4.04, 95% CI ‐5.44 to ‐2.64, P < 0.00001, I² = 56%). This may have been because patients received different types of surgery in these studies.

Length of intensive care unit (ICU) stay

We analysed 22 studies (see Characteristics of included studies) which reported the length of stay in ICU. The data were combined using the fixed‐effect model and revealed slightly prolonged ICU stay in the glutamine supplemented group (MD 0.18, 95% CI 0.07 to 0.29, P = 0.002, I² = 11%) (Analysis 1.5). Using the random‐effects model there was no significant difference in ICU days between the treatment and control groups (MD ‐0.12, 95% CI ‐0.62 to 0.37, P = 0.63, I² = 11%). The funnel plot for this meta‐analysis was symmetrical.

Of the 22 studies, eight studies (Andrews 2011; Conejero 2002; Déchelotte 2006; Grau 2011; Griffiths 1997; Hall 2003; Heyland 2013; Pérez‐Bárcena 2010) reported the days in ICU as the median (range) or median (interquartile range). They were changed into mean (standard deviation) as described above (Measures of treatment effect). A sensitivity analysis, excluding these studies by using the fixed‐effect model, reported the same result (MD 0.19, 95% CI 0.08 to 0.30, P = 0.001, I² = 22%). A sensitivity analysis including seven low risk of bias studies (Andrews 2011; Déchelotte 2006; Grau 2011; Griffiths 1997; Hall 2003; Heyland 2013; Çekmen 2011) and using the fixed‐effect model or the random‐effects model showed that glutamine supplementation had no effect on reducing the length of ICU stay (MD ‐0.54, 95% CI ‐1.48 to 0.4, P = 0.26, I² = 25%; MD ‐0.6, 95% CI ‐1.98 to 0.79, P = 0.40, I² = 25%, respectively).

Days on mechanical ventilation

Eleven studies with 455 patients (Chai 2006; Cucereanu‐Badica 2013; Eroglu 2009; Estívariz 2008; Fuentes‐Orozco 2004; Fuentes‐Orozco 2008; Garrel 2003; Heyland 2013; Pérez‐Bárcena 2008; Pérez‐Bárcena 2010; Zhang 2007) reported this outcome as the mean (standard deviation). The combined MD of these studies was ‐1.54 days (95% CI ‐2.44 to ‐0.65, P = 0.0008, I² = 0%) suggesting that glutamine supplementation reduced the number of days on mechanical ventilation in critically ill and surgical patients. In three studies (Déchelotte 2006; Grau 2011; Heyland 2013) the original data were reported as the median (interquartile ranges) or median (range). These studies were included in the meta‐analysis after they were converted into the mean (standard deviation) and the combined MD of all 14 studies was ‐0.69 days (95% CI ‐1.37 to ‐0.02, P = 0.04, I² = 18%) (Analysis 1.6). The funnel plot for this outcome was symmetrical, suggesting no evidence of publication bias for this outcome.

Sensitivity analysis including three low risk of bias studies (Déchelotte 2006; Grau 2011; Heyland 2013) showed that glutamine supplementation did not reduce the number of days on mechanical ventilation (MD 0.37, 95% CI ‐0.64 to 1.38, P = 0.47, I² = 32%).

Side effects

Eleven studies (Estívariz 2008; Çekmen 2011; Goeters 2002; Grau 2011; Lin 2002; Morlion 1998; Neri 2001; Pérez‐Bárcena 2008; Pérez‐Bárcena 2010; Yao 2007; Zhu 2000) reported that glutamine supplementation was well tolerated and no adverse events related to the administration of glutamine were observed. In two large sample, clinical trials (Andrews 2011; Heyland 2013) with 856 patients the serious adverse events outcome was reported and the pooled analysis of these two studies revealed that there was no significant difference in the risk of serious adverse events between the intervention and control groups (RR 1.40, 95% CI 0.77 to 2.57, P = 0.27, I² = 0%) (Analysis 1.7). In one trial (Heyland 2013) it was also found that the frequency of high urea levels (> 50 mmol/L) was higher among patients who received glutamine than for those that did not (13.4% versus 4.0%, P < 0.001). However, meta‐analysis of biological markers of liver and kidney function, such as alanine aminotransferase (Analysis 1.8), aspartate aminotransferase (Analysis 1.9) and creatinine (Analysis 1.10), did not find any difference between the groups during the treatment period. A further three studies reported on side effects. Two studies (Conejero 2002; Hall 2003) that supplemented glutamine enterally in critical ill patients reported gastrointestinal complications. The incidence of gastrointestinal complications was similar between the two groups. One study (Déchelotte 2006) of parenterally supplemented glutamine for ICU patients found that the number of adverse events per patient was significantly lower in the alanyl‐glutamine group than in the placebo group (2.0 versus 2.8, P < 0 .01).

Quality of life

This outcome was recorded in four studies. One study (Griffiths 1997) used the Whiston Health Questionnaire (Jones 1993) to assess patients' pre‐morbid health status and found that the morbidity of the participants at six months was similar between the glutamine supplemented and control groups. One study (Hall 2003) described the functional status of surviving patients at six months, and no differences were detected between the two groups. In one study (Powell‐Tuck 1999) patients were asked to score their mood, sleep, energy, appetite, pain, and mobilization on a 10‐point scale with 0 = 'perfect' and 10 = 'as bad as it can be'. No statistical difference was found between the intervention and control groups at the end of the treatment. In another study (Andrews 2011) quality of life measures (SF‐12 and EQ‐5D questionnaires) were collected at three and six months by post; the results have not been published.