The ability of perfusion index (PI) for predicting fluid responsiveness has been extensively studied with conflicting results. This meta-analysis evaluated the value of variation of PI (ΔPI) for predicting fluid responsiveness in adult patients undergoing preload challenge tests.

PubMed, Embase, and Cochrane Library were searched up to 20 ^th July 2026. Pooled sensitivity, specificity, diagnostic odds ratio (DOR), and area under the summary receiver operator characteristic curve (AUROC) were calculated. Q test and I ² statistics were used for study heterogeneity and publication bias was assessed by Deeks’ funnel plot asymmetry test. Subgroup analysis was performed for the potential causes of heterogeneity arising from varied patient characteristics.

Overall, nine studies including 492 preload challenge tests were administered to 419 patients in this review, of whom 52% were fluid responsive. The cutoff values of ΔPI were increased between 1.024% and 33% or decreased between -7% and -40%. Pooled sensitivity and specificity were 0.79 (95% CI: 0.71–0.85) and 0.82 (95% CI: 0.75–0.87), respectively. DOR was 17 (95% CI: 10–30) and AUROC was 0.88 (95% CI: 0.85–0.90), indicating good diagnostic accuracy. Heterogeneity was low ( I ²  = 28.95%, P  =  0.124). Subgroup analysis showed that ΔPI performed consistently across ICU and operating room settings, regardless of preload challenge type, PI device, or ventilation status.

Our meta-analysis suggests that ΔPI may serve as a promising adjunct non-invasive marker for predicting fluid responsiveness during preload challenge tests, but the current evidence base remains limited, heterogeneous, and at risk of bias.