Changes in stroke volume (deltaSV) induced by a lung recruitment manoeuvre (LRM) have been shown to accurately predict fluid responsiveness during protective mechanical ventilation. Cardiac output monitors are used in a limited number of surgical patients. In contrast, all patients are monitored with a pulse oximeter, that may enable the continuous monitoring of a peripheral perfusion index (PI). We postulated that changes in PI (deltaPI) may reflect deltaSV during brief modifications of cardiac preload. We studied 47 patients undergoing neurosurgery and ventilated with a tidal volume of 6-8 ml/kg. All patients were monitored with a pulse contour system enabling the continuous monitoring of SV and with a pulse oximeter enabling the continuous monitoring of PI. LRMs were performed by increasing airway pressure up to 30 cmH20 for 30 s. Fluid loads (250 ml of saline 0.9% in 10 min) were performed only in patients who experienced a deltaSV \textgreater 30% during LRMs (potential fluid responders). LRMs induced a 26% decrease in SV (p \textless 0.05) and a 27% decrease in PI (p \textless 0.05). We observed a fair relationship between deltaPI and deltaSV (r2 = 0.34). A deltaPI ≥ 26% predicted a deltaSV \textgreater 30% with a sensitivity of 83% and a specificity of 78% (AUC = 0.84, 95%CI 0.71-0.93). 24 patients experienced a deltaSV \textgreater 30% and subsequently received fluid. Fluid loads induced a 16% increase in SV and a 17% increase in PI, but fluid-induced deltaPI and deltaSV were weakly correlated (r2 = 0.19). In neurosurgical patients, we conclude that deltaPI may be used as a surrogate for deltaSV during LRMs but not during fluid loading.< Réduire