Résumé : Background: Continuous veno-venous hemofiltration (CVVH) can be used to reduce fluid overload and tissue edema, but excessive fluid removal may impair tissue perfusion. Skin blood flow (SBF) alters rapidly in shock, so its measurement may be useful to help monitor tissue perfusion. Methods: In a prospective, observational study in a 35-bed department of intensive care, all patients with shock who required fluid removal with CVVH were considered for inclusion. SBF was measured on the index finger using skin laser Doppler (Periflux 5000, Perimed, Järfälla, Sweden) for 3 min at baseline (before starting fluid removal, T0), and 1, 3 and 6 h after starting fluid removal. The same fluid removal rate was maintained throughout the study period. Patients were grouped according to absence (Group A) or presence (Group B) of altered tissue perfusion, defined as a 10% increase in blood lactate from T0 to T6 with the T6 lactate ≥ 1.5 mmol/l. Receiver operating characteristic curves were constructed and areas under the curve (AUROC) calculated to identify variables predictive of altered tissue perfusion. Data are reported as medians [25th–75th percentiles]. Results: We studied 42 patients (31 septic shock, 11 cardiogenic shock); median SOFA score at inclusion was 9 [8–12]. At T0, there were no significant differences in hemodynamic variables, norepinephrine dose, lactate concentration, ScvO2 or ultrafiltration rate between groups A and B. Cardiac index and MAP did not change over time, but SBF decreased in both groups (p < 0.05) throughout the study period. The baseline SBF was lower (58[35–118] vs 119[57–178] perfusion units [PU], p = 0.03) and the decrease in SBF from T0 to T1 (ΔSBF%) higher (53[39–63] vs 21[12–24]%, p = 0.01) in group B than in group A. Baseline SBF and ΔSBF% predicted altered tissue perfusion with AUROCs of 0.83 and 0.96, respectively, with cut-offs for SBF of ≤ 57 PU (sensitivity 78%, specificity 87%) and ∆SBF% of ≥ 45% (sensitivity 92%, specificity 99%). Conclusion: Baseline SBF and its early reduction after initiation of fluid removal using CVVH can predict worsened tissue perfusion, reflected by an increase in blood lactate levels.