Predicting Fluid Responsiveness in Mechanically Ventilated Adults: An Umbrella Review of Diagnostic Accuracy Meta-Analyses

Le résumé fourni par la source

BACKGROUND: Dynamic indices and maneuvers are widely used to predict fluid responsiveness in mechanically ventilated patients, yet their relative performance remains uncertain. METHODS: We conducted an umbrella review of diagnostic accuracy meta-analyses published up to August 2025. Eligible reviews evaluated pulse pressure variation (PPV), stroke volume variation (SVV), passive leg raising (PLR), end-expiratory occlusion test (EEOT), tidal volume challenge (TVC), pleth variability index (PVI), or respiratory variation in inferior vena cava diameter (ΔIVC) in mechanically ventilated adults. Redundancy across reviews was assessed using the Corrected Covered Area (CCA), with pairwise overlap explored using Jaccard coefficients. The primary umbrella-level comparative metric was pooled area under the receiver operating characteristic curve (AUC) as reported by the included meta-analyses. Pooled sensitivity, specificity, and hierarchical ROC information were extracted when available. Secondary analyses included cumulative meta-analysis, subgroup and meta-regression analyses, prediction intervals, and assessment of small-study effects/publication bias. RESULTS: Seven meta-analyses were included, synthesizing 123 primary diagnostic studies and 10,300 patients. Global CCA was approximately 5.5% to 5.6%, consistent with borderline slight-to-moderate overlap across the included meta-analyses. TVC showed the highest comparative diagnostic performance (AUC 0.96, 95% CI, 0.94-0.97), followed by EEOT (AUC 0.95, 95% CI, 0.92-0.96). Passive leg raising (PLR) achieved good accuracy (AUC 0.91, 95% CI, 0.88-0.93), whereas pulse pressure variation (PPV) and stroke volume variation (SVV) showed only moderate performance (AUC 0.87-0.88). Pleth variability index (PVI) and inferior vena cava variation (ΔIVC) were the least reliable (AUC 0.82-0.83). Subgroup and moderator analyses indicated that tidal volume was the main effect modifier: TVC and EEOT remained comparatively robust. CONCLUSIONS: Among dynamic predictors of fluid responsiveness, TVC and EEOT appear to be the most accurate and robust indices. PLR retains intermediate utility, PPV and SVV show moderate reliability, whereas PVI and ΔIVC should not be used in isolation.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.