The effects of higher versus lower protein delivery in critically ill patients: an updated systematic review and meta-analysis of randomized controlled trials with trial sequential analysis

Abstract

Background: A recent large multicentre trial found no difference in clinical outcomes but identified a possibility of increased mortality rates in patients with acute kidney injury (AKI) receiving higher protein. These alarming findings highlighted the urgent need to conduct an updated systematic review and meta-analysis to inform clinical practice.

Methods: From personal files, citation searching, and three databases searched up to 29-5-2023, we included randomized controlled trials (RCTs) of adult critically ill patients that compared higher vs lower protein delivery with similar energy delivery between groups and reported clinical and/or patient-centred outcomes. We conducted random-effect meta-analyses and subsequently trial sequential analyses (TSA) to control for type-1 and type-2 errors. The main subgroup analysis investigated studies with and without combined early physical rehabilitation intervention. A subgroup analysis of AKI vs no/not known AKI was also conducted.

Results: Twenty-three RCTs (n = 3303) with protein delivery of 1.49 ± 0.48 vs 0.92 ± 0.30 g/kg/d were included. Higher protein delivery was not associated with overall mortality (risk ratio [RR]: 0.99, 95% confidence interval [CI] 0.88-1.11; I² = 0%; 21 studies; low certainty) and other clinical outcomes. In 2 small studies, higher protein combined with early physical rehabilitation showed a trend towards improved self-reported quality-of-life physical function measurements at day-90 (standardized mean difference 0.40, 95% CI - 0.04 to 0.84; I² = 30%). In the AKI subgroup, higher protein delivery significantly increased mortality (RR 1.42, 95% CI 1.11-1.82; I² = 0%; 3 studies; confirmed by TSA with high certainty, and the number needed to harm is 7). Higher protein delivery also significantly increased serum urea (mean difference 2.31 mmol/L, 95% CI 1.64-2.97; I² = 0%; 7 studies).

Conclusion: Higher, compared with lower protein delivery, does not appear to affect clinical outcomes in general critically ill patients but may increase mortality rates in patients with AKI. Further investigation of the combined early physical rehabilitation intervention in non-AKI patients is warranted.

Prospero id: CRD42023441059.

Keywords: Critical illness; Physical rehabilitation; Protein; Systematic review.

Fig. 1

Meta-analysis of clinical outcomes. a Overall mortality (all patients), b Overall mortality (subgroup analysis of no/not known AKI vs AKI)*, c infectious complications (no change from previous meta-analysis), d ICU length of stay, e hospital length of stay, f duration of mechanical ventilation. AKI: acute kidney injury. *Note: b, c: AKI subgroup: mortality from Doig 2015 is 90-d mortality from their secondary publication [13]. Definitions: Singer 2007: AKI—50% decrease in GFR, a doubling of serum creatinine or an increase of creatinine to 3.5 mg/dL (309.4 umol/L); Doig 2015 (mortality of patients with kidney dysfunction or risk of progression of AKI from Doig 2015 is 90-d mortality from their secondary publication [13]): Baseline kidney dysfunction—creatinine at time of enrolment > 168 μmol/L (by Gordon Bernard’s “Brussels Table”), Risk of progression of AKI at enrolment—a rise in creatinine over the previous 24 h by at least 20% to over 120 μmol/L; Heyland 2023: AKI—patients who met the criteria of KDIGO: stage 1 is at least 26·52 μmol/L increase in serum creatinine from baseline within 48 h or 1·5–1·9 times baseline within 7 days, stage 2 is 2·0–2·9 times baseline within 7 days, or stage 3 is three times or more baseline within 7 days or increase to at least 353·6 μmol/L with an acute increase of more than 44·2 μmol/L. c: To ascertain the mortality count and total sample size for the no/not known AKI subgroup for Doig 2015 and Heyland 2023, the mortality count and total sample size of the AKI subgroup were subtracted from the overall mortality count and total sample size, respectively; mortality for Doig 2015 is 90-day mortality

Fig. 2

Meta-analysis of other outcomes. a Percentage of muscle change per week (no changes from previous meta-analysis), b handgrip strength, c discharge to rehabilitation facilities (no changes from previous meta-analysis), d self-reported quality of life physical function at day 90, e incidence of diarrhoea (no changes from previous meta-analysis). Note: b Fetterplace 2018: the best handgrip strength at awakening, ICU discharge, or day 15, Ferrie 2015: handgrip strength at day 7. Unable to analyse handgrip strength from Azevedo 2019 because unknown sample size for male and female. d The quality of life (QOL) outcomes reported by the studies were: Doig 2015: RAND-36 general health and physical function at day 90; Azevedo 2019: SF-36 physical component summary (PCS) score at 3 and 6 month; Badjatia 2010: fatigue, lower extremity mobility, and cognition outcomes based on the Neuro-QoL questionnaires administered on post-bleed day 90; Chapple 2020: EQ-5D-5L score for mobility, self-care, usual activities, pain/discomfort, anxiety/depression, and the result of the EQ-5D-5L visual analogue scale, all at day 90, Azevedo 2021: SF-36 physical component score at day 3 and 6 month (see Additional file 1: Table S9). The meta-analysis was performed for QOL results associated with physical function: RAND-36 physical function at day 90 (Doig 2015), SF-36 PCS score at 3 month (Azevedo 2019), Neuro-QoL lower extremity mobility on post-bleed day 90 (Badjatia 2010), EQ-5D-5L score for mobility at day 90 (Chapple 2020), and SF-36 physical component score at 3 month (Azevedo 2021). Higher EQ-5D-5L mobility score means worse performance; a negative is added to the mean score to reverse the direction of the results

Fig. 3

Trial Sequential Analysis of Clinical Outcomes. a Overall mortality in all patients (21 studies, n = 3125), b overall mortality in patients with acute kidney injury before protein intervention (3 studies, n = 428), c infectious complications (7 studies, n = 642), d intensive care unit length of stay (16 studies, n = 2516), e hospital length of stay (11 studies, n = 2130), f duration of mechanical ventilation (13 studies, n = 2360). TSA was analysed using DerSimonian and Laird random-effects model. The Z curve in blue measures the treatment effect (pooled relative risk). The parallel lines in green are the boundaries of conventional meta-analysis (alpha 5%), and the boundaries of benefit and harm are boundaries of conventional meta-analysis adjusted for between-trial heterogeneity and multiple statistical testing (TSA boundaries). A treatment effect outside the TSA boundaries of benefit/harm indicates reliable evidence for a treatment effect, and a treatment effect within the futility zone (the triangle between the parallel lines) indicates that there is reliable evidence of no treatment effect. DARIS: diversity adjusted required information size is the calculated optimum sample size for statistical inference, MID: minimally important difference, RRR: relative risk reduction, TSA: trial sequential analysis