Nutritional interventions in patients with heart failure and malnutrition: a systematic review and meta-analysis of randomized trials

REGISTRATION TIMING This record was created after the review was completed and is a retrospective registration. The review was not prospectively registered. The literature search was executed on 4 October 2026, and study selection, data extraction, risk-of-bias assessment and the meta-analyses were completed before this record was created. A later registration cannot establish that the decisions recorded here were made before the results were known. The pre-specified methods are reproduced in the protocol extract attached to this record, and every dated deviation from them is summarised below.

Objective

To determine, in adults with heart failure and malnutrition or at risk of malnutrition, whether a nutritional intervention changes all-cause mortality, the composite of death or heart failure hospitalisation, and heart failure hospitalisation, and to describe its effect on nutritional, functional and quality-of-life outcomes. ELIGIBILITY CRITERIA Population: adults with chronic or decompensated heart failure and malnutrition or at risk of malnutrition, in any care setting. The criterion is applied as malnutrition, nutritional risk or documented nutritional depletion (for example severe muscle depletion with a body mass index in the normal range). Intervention: any strategy intended to increase energy or protein intake or to reverse malnutrition, including oral nutritional supplements, food fortification or enrichment, protein supplementation, dietary counselling, medical nutrition therapy and enteral or parenteral nutrition. Comparator: usual care, standard dietary advice or placebo. Outcomes: co-primary - the composite of death or heart failure hospitalisation, and all-cause mortality. Secondary - heart failure hospitalisation, nutritional status, functional capacity, quality of life and body composition. Design: randomised controlled trials. Exclusions: isolated micronutrient supplementation (vitamin D, selenium, magnesium, thiamine, coenzyme Q10) in participants without documented malnutrition or nutritional risk. Micronutrients delivered as part of a broader energy- or protein-focused intervention were not excluded. INFORMATION SOURCES AND SEARCH DATE PubMed/MEDLINE (222 records), Cochrane CENTRAL (120) and LILACS/BVS (4); 346 records in total, searched on 4 October 2026 without a language restriction. Trial-registration records were retrieved through CENTRAL; no separate register search was run. The PubMed strategy combined the condition, a malnutrition block and an intervention block with a randomised-trial filter; the CENTRAL strategy used four blocks and the LILACS strategy three bilingual blocks. The strategies are reproduced in full in the attached protocol extract. STUDY SELECTION AND DATA COLLECTION Two reviewers screened titles and abstracts independently and screened full-text reports independently, resolving disagreements by discussion. Two reviewers extracted data independently. No automation tools were used and the authors of the included studies were not contacted; the analyses rest on published aggregate data. Of the 346 records, 57 duplicates and 34 trial-registration records without posted results were removed before screening; 255 were screened and 232 excluded, leaving 23 reports sought for retrieval, of which one could not be retrieved (Arutiunov 2009, published in Russian without a DOI). Of the 22 reports assessed for eligibility, 6 were excluded (1 non-nutritional intervention, 5 ineligible design), leaving 16 eligible reports; 5 were duplicate or redundant publications of trials already identified and were consolidated at the study level, giving 11 unique randomised trials. RISK OF BIAS AND CERTAINTY Risk of bias was assessed with ROBUST-RCT, item by item, without an aggregated score. The certainty of the body of evidence was assessed with GRADE across the five domains. SYNTHESIS Clinical outcomes: the risk ratio was computed from the 2x2 counts and pooled by Mantel-Haenszel for the common-effect model and by DerSimonian-Laird for the random-effects model, with the Knapp-Hartung standard-error adjustment reported alongside. Continuous outcomes: mean difference, with values at final follow-up and change from baseline permitted in the same analysis and the standard deviation of a change imputed where necessary with the correlation coefficient fixed at 0.5 (sensitivity analysis at 0.2 and 0.8). Software: Stata 16 or later. Heterogeneity was described with tau-squared and I-squared and investigated by subgroup, meta-regression and leave-one-out analyses. Small-study effects and publication bias were examined with contour-enhanced funnel plots, the Harbord and Egger tests and trim-and-fill. MULTIPLICITY The two co-primary outcomes carry the inferential weight; every other analysis is exploratory. No correction for multiplicity is applied. PRE-SPECIFIED POOLING THRESHOLD Continuous outcomes were pooled only where at least two trials shared the instrument, the unit, the time point and the dispersion. In execution only body mass index and albumin met the threshold; the remaining continuous outcomes are reported as single-trial estimates. DEVIATIONS FROM THE PLAN Five deviations are declared and dated: (1) the synthesis software was changed from R to Stata; (2) standardised mean differences were planned for functional, quality-of-life and body-composition outcomes, but only body mass index and albumin met the pooling threshold and were pooled as mean differences; (3) PICNIC, which stopped early for benefit, was included in the primary analysis and examined in a leave-one-out sensitivity analysis; (4) the 30-day mortality analysis of EFFORT-HF is exploratory and post hoc; (5) the meta-regression on follow-up duration is exploratory.

Funding

AND COMPETING INTERESTS No funding was received and the authors declare no competing interests. DATA AND CODE The aggregate 2x2 data and the analytical code that reproduce every estimate are available from the corresponding author.

Results

The review is complete; eleven randomised trials were included, of which five reported clinical events. All-cause mortality (5 trials, 1,131 patients, 286 deaths) gave a risk ratio of 0.790 (0.646 to 0.965) under the common- effect model and 0.770 (0.562 to 1.054) under the random-effects model, rising to 0.770 (0.492 to 1.203) with the Knapp-Hartung adjustment; I-squared was 33.8 per cent. The composite of death or heart failure hospitalisation (3 trials, 410 patients) gave 0.577 (0.424 to 0.785) common-effect and 0.574 (0.212 to 1.552) Knapp-Hartung, and heart failure hospitalisation 0.537 (0.353 to 0.816) and 0.531 (0.099 to 2.840). Certainty of evidence is very low for all three clinical outcomes, imprecision being the decisive domain.