Oral nutritional interventions in hospitalised older people at nutritional risk: a network meta-analysis of...

Malnutrition affects 35% to 64% of hospitalised older people, and is associated with adverse health outcomes such as disease complications and hospital readmission. Identifying effective nutritional interventions is essential to improve clinical outcomes and reduce healthcare costs in this population.

To evaluate the effects of various nutritional interventions, compared with either a control group (standard care or placebo) or each other, on patient-relevant outcomes in hospitalised older people at risk of or with established malnutrition, and to rank the effects of these different interventions using network meta-analysis (NMA) based on individual participant data (IPD).

We searched CENTRAL, MEDLINE, five other databases, and two trial registries to 2 July 2024, and checked the reference lists of included studies and relevant systematic reviews.

We included older people (≥ 65 years) hospitalised for different acute conditions at risk of or with malnutrition enrolled in randomised controlled trials (RCTs) comparing oral nutritional interventions with control or each other. For RCTs that met our inclusion criteria, either fully or partially, we requested IPD from the study authors. If we did not receive a response or IPD were unavailable, we used published aggregated data. We excluded RCTs that only partially met the eligibility criteria if neither IPD nor sufficient aggregated data were obtainable.

Critical outcomes were all-cause mortality, serious adverse events (SAEs), and functional status (e.g. activities of daily living). Important outcomes were health-related quality of life (HRQoL), length of hospital stay (LOS), body weight, and fat-free mass. The main outcome assessment time point was at hospital discharge or 30 days after randomisation.

We used the Cochrane risk of bias 2 (RoB 2) tool.

For each outcome, we first analysed IPD within each study. Second, we pooled results in an NMA which also included the aggregated data from RCTs without available IPD. We performed random-effects NMAs based on the frequentist approach and ranked treatments by P-scores. We rated the certainty of evidence using the GRADE approach.

We included 21 RCTs (72 reports; 12 RCTs with IPD) with 3309 older participants (mean age ranged from 75 to 85 years; 1863 participants with IPD) with different acute conditions. Interventions included the provision of additional protein (three studies), energy supplements (two studies), oral nutritional supplements (ONS; eight studies), individualised feeding support (two studies), and comprehensive individualised nutritional care (eight studies). In all but two RCTs, interventions were compared to control (standard care with or without a placebo). We judged 16.1% of outcome assessments to be at low risk of bias and 16.8% at high risk.

ONS may reduce all-cause mortality (risk ratio (RR) 0.46, 95% confidence interval (CI) 0.25 to 0.84; absolute risk difference 57 fewer deaths per 1000 people, 95% CI 79 fewer to 17 fewer; low-certainty evidence) compared to control, while comprehensive individualised nutritional care may show little to no effect (RR 0.98, 95% CI 0.55 to 1.73; 1 fewer per 1000 people, 95% CI 26 fewer to 46 more; low-certainty evidence). For all other treatment comparisons, the evidence is very uncertain (NMA with 13 RCTs, 2728 participants; Q between designs: not applicable (NA)). ONS may reduce SAEs compared to control (RR 0.56, 95% CI 0.32 to 0.95; 84 fewer SAEs per 1000 people, 95% CI 131 fewer to 10 fewer; Q between designs: Q 1.95, df 2, P = 0.3772; low-certainty evidence). For all other treatment comparisons, the evidence is very uncertain (NMA with 14 RCTs, 2184 participants). Comprehensive individualised nutritional care may make little to no difference in activities of daily living compared to control (standardised mean difference (SMD) 0.06, 95% CI -0.08 to 0.20; low-certainty evidence) and ONS compared to energy supplements (SMD -0.15, 95% CI -0.53 to 0.23; low-certainty evidence). For all other treatment comparisons, the evidence is very uncertain (NMA with 5 RCTs, 1128 participants; Q between designs: NA). Energy supplements probably make little to no difference in HRQoL compared with ONS (mean difference (MD) 0.01, 95% CI -0.06 to 0.08; Q between designs: NA; moderate-certainty evidence). All other comparisons of different nutritional interventions may make little to no difference to HRQoL (NMA with 3 RCTs, 1513 participants). The provision of additional protein, energy supplements, ONS, and comprehensive individualised nutritional care may make little to no difference in LOS compared to control (18 RCTs, 3013 participants; Q between designs: Q 2.86, df 3, P = 0.4145). Body weight (16 RCTs, 2114 participants; Q between designs: Q 2.03, df 3, P = 0.5655) may increase with ONS when compared to control (MD 0.9 kg, 95% CI 0.37 to 1.42) or comprehensive individualised nutritional care (MD 1.00 kg, 95% CI 0.12 to 1.87), but the evidence is very uncertain. Energy supplements and ONS probably have similar effects on body weight (MD 0.11 kg, 95% CI -0.85 to 0.63; moderate-certainty evidence). For fat-free mass, no meta-analysis was possible. One RCT (102 participants) compared ONS with energy supplements and found little or no difference between groups (MD 0.13 kg, 95% CI -0.63 to 0.90; low-certainty evidence), while evidence regarding the effects of additional protein compared with control was very uncertain (1 RCT, 19 participants). Rankings of treatments by P-scores were not consistent across outcomes.

In older hospitalised people at risk of or with malnutrition, oral nutritional supplements may reduce mortality and SAEs compared to control 30 days after randomisation. For other outcomes, there may be little or no differences in results. Overall, the evidence was of low to very low certainty, primarily due to a limited number of studies and participants per comparison. The comparison of treatment effects across outcomes was constrained by variations in network structure. When interpreting the results, the heterogeneity of the population in terms of acute and chronic conditions needs to be considered. To improve certainty, adequately powered studies with robust methodologies should compare interventions with controls as well as against each other.

The German Federal Ministry of Education and Research funded this work (grant number: 01KG2102).

Protocol (2022) doi.org/10.1002/14651858.CD015468.