In this blog post, we'll explore the various components of the trial and seek reasons for why glutamine supplementation failed to be effective.

Many nutritional additives have been tested in the critical care setting. To date, almost all such pharmaconutrients have failed to convincingly demonstrate benefit in large randomised controlled trials, whether delivered intentionally as pharmaconutrients or not. These include selenium,² antioxidants,³ omega-3 fatty acids,⁴ vitamin D,⁵ and vitamin C.⁶

A large body of evidence supports the premise that glutamine has beneficial biological effects in the setting of burns. Firstly, critically ill patients have lower glutamine plasma levels than non-critically ill patients, and this reduction is associated with poorer outcomes.⁷ This is due to a combination of increased demand for a conditionally essential amino acid in critical illness⁸ and exudative losses from burns.⁹ Secondly, glutamine levels fall rapidly in the setting of an acute burn.¹⁰ Thirdly, many acute processes require glutamine as a fuel,⁸ with the immune system dependent on it for adequate lymphocyte and macrophage function, while enterocytes require it to maintain gut activity.¹¹ However, none of these observations confirm or refute that glutamine supplementation is beneficial. It may simply be a biomarker of severity of illness and not lie on a causal pathophysiological pathway. The literature is replete with critical care trials restoring various plasma levels to normal, but without positively impacting outcomes, such as albumin,^12,13 vitamin D,⁵ haemoglobin,¹⁴ and fibrinogen.¹⁵

Physiological variables were not presented in the trial publication, either at baseline or post randomisation, making it difficult to determine whether glutamine supplementation exerted a biological effect or not. However, it is worth noting that the adequacy of nutrition was similar in both the glutamine and placebo groups, suggesting little impact on gastrointestinal tract function. Time to start of enteral nutrition (0.7 vs 0.8 days), duration of enteral (11.5 vs 11.3 days) and parenteral (9.0 vs 9.0 days) feeding, calorie adequacy (74% vs 72%) and protein adequacy (75% vs 72%) were approximately equal, as were gastrointestinal complications. Consistent with this, there was no difference in gram-negative bacteraemias, or infectious complications. Thus, the two main physiological systems, the immune and gastrointestinal systems, expected to benefit most did not appear to be impacted by glutamine supplementation. The global measure of mortality was also comparable between groups. Although all three are crude and insensitive measures, there is no signal of an obvious biological impact or therapeutic efficacy.

At present, no mechanistic substudies appear to have been published to illuminate this discussion.

RE-ENERGIZE was a large international, multi-centre randomised control trial. It ran in 14 countries and included 54 centres. While the wide geographic inclusion improves external validity, it forces consideration of its effect on internal validity, especially given the summative nature multiple interventions delivered outside the protocol could have.

Globalisation bias could describe this heterogeneity. The American Society for Parenteral and Enteral Nutrition (ASPEN)¹⁶ and the European Society for Clinical Nutrition and Metabolism (ESPEN)¹⁷ provide different guidance in enteral- versus parenteral nutrition recommendations: ASPEN recommends seven, while ESPEN recommends three days before starting supplemental parenteral nutrition in critically ill patients. This potential problem was avoided because the study centers underwent review before entering the study to standardize the nutrition and micro-nutrition protocols. The centers included in this trial were, however, otherwise allowed to proceed with burns care at their own discretion, which can create further heterogeneity.

Multiple variances in practice can occur across the world, including the differing use of fluid resuscitation protocols, fluid volume administered, fluid type, resuscitation end points, timing of burns excision, antibiotic use, sedation and analgesia use. This is reflected in the use of numerous burns protocols across the world.¹⁸ Characterising the severity of a burn varies across centres, with specialised burns centres being more accurate.¹⁹ Other differences include the frequency of dressing change and the type of dressing used.²⁰ With the paucity of burns centres, due to their highly specialised nature, most patients are likely to have initially presented to a non-burns hospital.

However, despite these variances across different locations, it is expected they should largely balance out. Randomisation functions to balance risk between groups, by broadly equalising key variables, both measured and unmeasured. Between-group balance in measured potential confounding variables, such as total body surface area burn, allows the inference that unmeasured variables are likely to be balanced between groups also, and thus that the risk of study outcomes (such as mortality) is evenly distributed between groups at baseline. 

There are two key design differences between the RE-ENERGIZE trial and most of the preceding trials investigating glutamine supplementation in patients with severe burns.

The first is the size of the trials, with earlier trials being small, enroling an average of just 42 patients each. In contrast, RE-ENERGISE recruited 1209 patients, over three times the combined total of the other 9 burns trials.^{10,21,22,23,24,25,26,27,28} Smaller studies finding benefit or harm may be more likely to be reported and/or published than smaller inconclusive studies, creating both reporting and  publication bias. Smaller sample sizes may also overestimate the effect size due to random high outliers, leading to false-positive results. Smaller trials (often single-centre) typically have stricter inclusion criteria to maintain patient homogeneity. Fewer patients also means less opportunity for more divergent care, especially in a single-centre setting. Thus, small trials often address the question "in these carefully controlled and specific experimental conditions, does the intervention alter the likelihood of an outcome or outcomes?". While this can enhance internal validity, it may limit the external validity or generalizability of the results to the broader population. Small trials are more susceptible to various biases. For instance, allocation bias or implementation bias can have a proportionally larger impact in a small trial. Additionally, small studies might not adequately randomise confounding variables by chance, leading to skewed results. Smaller trials may not provide sufficient evidence for replication, as the results might be specific to the particular cohort recruited or setting. Larger trials, on the other hand, are more likely to produce replicable results due to their more diverse and representative samples.²⁹ 

The second design comparison is the multi-centre nature of RE-ENERGIZE, which contrasts with 8 of the other 9 burns trials being single-centre (and all recruiting 60 patients or less).^{10,21,22,23,24,25,26,28} Many of these issues are common to both small and single-centre trials, for the obvious reason that both designs suffer from limited sample sizes.  Single-center trials are more prone to biases compared to multi-center trials.³⁰ These biases include local effect bias, selection and performance bias, detection and reporting bias, analysis and attrition bias, concomitant therapy bias, and publication bias. These biases can lead to an overestimation of the treatment effect in single-centre trials which may not be replicable in the more varied and rigorous setting of multi-centre trials. Single-centre trials report larger treatment effects than multi-center trials, even after adjusting for sample size, risk of bias within trials, and accounting for publication bias.³¹ Their results may be highly contextual and specific to the unique environment of the single center, which may not be generalisable to other settings or populations, as is often the case in multi-center trials.²⁹ The single-centre trials reporting benefit from glutamine administration in patients with burns were likely type 1 errors.³²

A major methodological consideration with this trial was the modification to the protocol whilst the study was enroling. Recruitment was slow and after four years just 690 patients had been enroled out of a planned 2,700.  For this reason, the primary outcome (six-month mortality) and secondary outcome (time to hospital discharge) were switched.

From the perspective of understanding why RE-ENERGIZE reported a null result, this protocol modification raises possible answers.

Firstly, the primary outcome is typically the one with the most statistical power and changing it with a less-powered secondary outcome can compromise the study’s validity. RE-ENERGIZE was powered for a 25% relative reduction in six-month mortality, from 15% to 11.5%, with 80% power and a two-sided alpha of 0.05%. This required 1,273 patients per arm, rounded to 2,700 patients for a potential 5% attrition rate. The originally published protocol from 2017 describes how this 2,700 patients would also have 80% power to identify a daily hazard rate of discharge among survivors of 1.16 for patients receiving glutamine. However, at the time of the protocol change, the blinded dataset was analysed to determine the pooled event rate to that point. It was estimated, based on this data, that if the intervention group had a 20% relative reduction in 90-day mortality, (from 15.56% to 12.44%), in addition to a 20% relative increase in the daily rate of discharge amongst those surviving to 90 days, then 1200 patients would have 80% power to detect a difference in time to discharge at a two-sided alpha of 0.05.

Thus, while the statistical plan changed, there was a re-powering based on blinded data. The competing risk of death was also accounted for. At no point in the 90 day period that hospital outcomes were measured did the glutamine group ever have a higher cumulative incidence of discharge alive from hospital in comparison to the control group. While the confidence interval includes the possibility of a shorter median time to hospital discharge (subdistribution hazard ratio for discharge alive, 0.91; 95% CI, 0.80 to 1.04; P=0.17), this seems unlikely. As such, the probability this is an underpowered trial, as a result of the protocol change, appears remote.

While there was no ultimate direct impact from the power calculation in RE-ENERGIZE, the trial indirectly suffered from the multiple smaller studies preceding it, all with limited sample sizes, meaning any prior effect size seen may not have been a true finding.³³ However, multiple smaller studies reporting benefit do require larger confirmatory trials.

The trialists sought to optimise the enroled population by only including severely injured patients who were felt to be the most likely to benefit from glutamine supplementation. Patients had either second or third-degree burns and an expectation for skin grafting. As patient age is an important modifier of outcome, the required burn size differed by age, being >20% for those aged 18 to 39 years, > 15% for those 40 to 59 years and > 10% for those aged 60 years and above. Those unable to receive enteral glutamine were excluded, as were moribund patients.

Could the recruited cohort have been sufficiently different from those in earlier trials reporting benefit with glutamine to explain why the intervention was ineffective in RE-ENERGISE?

It is extremely difficult to compare populations across small, single centre trials, especially in patients with burns. Initial resuscitation usually begins at non-burns centres, where estimations of the total body surface area,  burn depth, need for escharotomy, intubation and initial fluid resuscitation may be inaccurate.¹⁹ These decisions have a major impact on patient outcomes.³⁴ As such, local treatment effects may be pronounced, as may regional and national effects, reflecting critical care in general.³⁵

In RE-ENERGIZE, 0.5 g/kg/day of glutamine was administered in between 3 and 6 divided doses, until 7 days after the last skin graft procedure and up to a total of three months after admission. The mean intervention period for the 1209 patients was 26 days, with 91% of the doses administered. Thus, the intervention was probably adequately delivered.

In contrast to the REDOXS trial,³⁶ which administered 0.35 g/kg/day of glutamine intravenously plus 30 g/day enterally, and reported harm from this approach, RE-ENERGIZE investigated a lower dose of 0.5 g/kg/day, delivered enterally. This was also in the setting of probably much larger loses of glutamine due to the presence of burns exudates. Glutamine monotherapy, in doses ranging from 0.2 g/kg/day to 0.5 g/kg/day, has not been found to reduce mortality in critically ill patients, although its effects in subpopulations in intensive care are less clear.³⁷ ASPEN published a position paper in 2011 where a dose recommendation could not be made, but it was estimated to be higher than 0.2 g/kg/day and "probably closer to 0.5 g/kg/day”, providing further justification of the chosen dose.³⁸ In 10 randomised controlled trials of glutamine in critically ill patients with burns, 7 used a dose of 0.5 g/kg/day^{1,21,23,25,26,27,28} and 2 a dose of 0.35 g/kg/day,^10,24 either enterally or intravenously. One trial used 26 g/day.²² Thus, the dose tested appeared adequate, neither too high nor too low.

While this trial administered glutamine enterally, 5 small randomised controlled trials have used the intravenous route.^{21,24,26,27,28} A meta analysis including these 5 trials did not find a mortality reduction with intravenous administration (RR, 0.60; 95% CI, 0.21 to 1.72; P=0.34),³² suggesting parenteral administration would not have changed the result obtained.

The placebo consisted of a maltodextran-based solution. Maltodextran is a highly processed carbohydrate which rapidly undergoes hydrolysis to glucose. The additional calorific load was felt to be trivial and similar to that of the glutamine administered to the intervention group.³⁹ The maltodextran placebo had no metabolic effects.³⁹  Interestingly, the placebo did not contain protein as an adequate amount was delivered in the daily feed. Also, the protein load contributed by glutamine was not included in the total dose of protein delivered to each patient in the intervention group, to maintain blinding of the dietitian. While both groups were intended to receive adequate protein from their daily feed, the additional protein load from glutamine left the intervention group vulnerable to receive a higher protein intake than the control group. The Dipeptivan (glutamine) product summary sheet⁴⁰ from Fresenius Kabi specifically states the protein load from supplemental glutamine administration should be considered when calculating total daily protein dosing, although glutamine administration in RE-ENERGIZE was clearly in the setting of a randomised controlled trial rather than usual clinical care.

This raises the question as to whether the additional protein load from glutamine supplementation could have produced a negative effect. Consistent with this speculation, the median (IQR) urea level was higher in the glutamine group {8.7 mmol/L (5.9 to 15.6) vs. 6.8 (4.8 to 12.2)}. The EFFORT trial⁴¹ recently reported harm from a high protein nutritional strategy - ≥ 2.2 g/kg/day vs ≤ 1.2 g/kg/day. Opposing this supposition is the fact that protein adequacy was similar in both groups at ~73%, suggesting neither group received an excess of protein. Although nutritional practices were standardised between groups, and internationally across centres, it was not reported either what these were or what dose of protein was actually delivered in the trial.

This finding of a higher urea level in the glutamine group was also seen in the REDOXS trial,³⁶ investigating glutamine in a broader critically ill population (urea > 50 mmol/L; 13.4% vs. 4.0%, P<0.001). It was speculated in a post-hoc analysis of REDOXS⁴² that a metabolite of glutamine may accumulate in renal injury contributing to the harm seen in that trial. Interestingly, this signal of harm was abated in those receiving renal replacement therapy, a finding also seen in the EFFORT trial. As such, the finding of an increased urea may be a manifestation of glutamine-induced renal injury rather than a direct protein-load effect, although it is possible that both processes co-exist.

Randomisation functioned as expected to produce two broadly similar groups at baseline, with no major differences in their observed traits. With effective blinded group allocation, study drug administration, outcome reporting and analysis, the trial appears to have been executed as planned.

Of the included 1209 patients, 63 suffered severe adverse events. None were deemed related, except one episode of kidney failure, which was considered “possibly” related. This implies an accumulation of adverse events was not the mechanism producing the null result.

In contrast, in the REDOXS trial³⁶ in a general ICU population, glutamine supplementation resulted in increased mortality at 14 days (25.7% vs 21.3%; P=0.07), 28 days (32.4% vs 27.2%; P=0.05), in-hospital (37.2% vs 31.0%; P=0.02), and 6 months (43.7 % vs 37.2%; P=0.02). Amongst 13 subgroup results, 12 point estimates suggested harm with glutamine.

Viewing both of these large trials, REDOXS & RE-ENERGIZE, which together dwarf all other trials of glutamine in critical care, there is, at best, no signal of benefit from this intervention, and possibly one of harm.

Glutamine has been extensively studied in critically ill patients. The largest trial in a general critically ill population reported harm. Now, the largest trial in burns patients has reported a lack of clinical effect, either beneficial or harmful. Reviewing the trial, no obvious bias accounts for this result. Based on the currently available evidence, the clinical implications appear straightforward - the additional supplementation of glutamine, as dosed in the RE-ENERGIZE trial, in patients with burns is ineffective.

Surprisingly, the latest updated guidance from the European Society for Clinical Nutrition and Metabolism, published in July 2023 and incorporating the RE-ENERGIZE trial results, continues to advocate for the use of glutamine supplementation in critically ill burns patients.⁴³ In contrast, the latest systematic review and meta analyses from the RE-ENERGIZE trialists reports a lack of benefit from glutamine supplementation in burns patients.³² The finding of benefit from glutamine supplementation was limited to small, single-centre trials, a theme consistent throughout critical care research. Furthermore, the most recent meta analysis investigating the efficacy of glutamine in a general critically ill population, including 2552 patients from 18 randomised controlled trials, also reported no mortality benefit, or reduction in ICU length of stay or infectious complications.⁴⁴

The global burden of burns is enormous.⁴⁵ Every year, over 7 million people sustain burn injuries globally. These injuries account for the loss of close to 18 million years of healthy life when adjusted for disability (DALYs), and lead to over a quarter of a million deaths. With the vast majority of this impact, exceeding 90%, suffered by people in low- and middle-income nations, the avoidance of an ineffective therapy in potentially resource-limited settings is helpful, as is the reinforced message to reduce or eliminate low value care.

A wider implication of the RE-ENERGIZE trial, for not just burns management or even critical care, but for the whole medical community, is the importance of correctly interpreting single-centre trials, especially small ones, even if replicated. Single-centre trials reporting benefit rarely survive replication in larger multi-centre environments,⁴⁶ for a multitude of reasons.⁴⁷ The costs of early clinical adoption of such trials,^48,49 even when alluring, can be substantial and take years to reverse.

RE-ENERGIZE was a large, international, multi-centre, blinded, randomised controlled trial evaluating glutamine supplementation in critically ill patients with severe burns. It was well designed and robustly executed. The results are solid and actionable. It is a sentinel warning to beware of overly interpreting small trials, even when multiple, and a triumph of science, where larger-scale iteration moves us closer to the truth.