119 PROTEIN PROVISION AND EXERCISE THERAPY TO IMPROVE OUTCOMES IN ICU-SURVIVORS 7 6-12 respectively. The ICU length of stay (ICU-LOS) varied between 5-21 days. Intervention strategies were heterogeneous between studies. For example, the target protein and energy requirements could be based onmeasurements (24-h urinary urea, indirect calorimetry) or equations (g/kg/day or kcal/kg/day) using different patient weights (actual or ideal body weight). In RCTs, the protein target ranges in the intervention group were 1.2-2.2g/kg/ day and 0.8-1.5g/kg/day in the comparator group. For NRSI, the range was between >0.7-≥1.2g/kg/day and <0.7-<1.0g/kg/day in the intervention and comparator group respectively. In addition, a combination of EN and PN was frequently used, in some studies enriched with protein supplements or including other sources of calories like propofol or dextrose. One study used exclusively PN as an intervention strategy.55 All interventions took place in the early phase of ICU-admission; no studies in the post-ICU phase were included. The study of Nakamura57 was the only included study with both protein provisionandcombinedprotein-exercise therapy reported. Active rehabilitation in this study consisted of exercise therapy (2x20min./day) with a physical therapist and nurse, combined with NEMS (1x20min./day). 7.3.3 Risk of bias A total of nine RCTs were judged for risk of bias (Appendix 2, figure S1).46,47,50,51,53– 55,57,58 Most studies did not reveal bias arising from the randomisation process. Six out of nine RCTs were judged with “some concerns”46,47,50,53,54,57 and three studies as “high”51,55,58 risk of bias. The direction of bias was mainly in favour of the intervention group, especially due to blinding difficulties and missing outcome data. Seven domains concerning the risk of bias inNRSI were assessed for six studies, only for mortality outcomes (Appendix 2, figure S2).48,49,52,56,59,60 All included NRSI were retrospective cohort studies with the most common biases came from confounding, selection of participants, and classification of interventions. The retrospective nature of the studies complicated the feasibility of measuring and controlling for all relevant variables. In addition, inadequate adjustment techniques for variables that correct for selection bias were present.48,59,60 Overall, five out of six NRSI studies were judged as “serious” risk of bias48,52,56,59,60 and one as “moderate” risk.49 7.3.4 Effects of protein provision on quality of life Two studies reported QoL outcomes for the effect of protein provision (Figure 2 and Appendix 2; table S1).50,57 There was no statistically significant difference in QoL between intervention and comparator at hospital discharge in the study by Nakamura57 (MD 4.3, CI [-1.6;10.2], SMD 0.39, CI [-0.14; 0.92]). Chapple et al50
RkJQdWJsaXNoZXIy MjY0ODMw