Local Warming to Improve Pulse Oximetry Accuracy in Critically Ill Patients

Overview

Pulse oximeters are widely used in intensive care to continuously estimate the oxygen level in the blood. However, pulse oximeter measurements may be less accurate when blood flow to the fingers is poor, which is common in critically ill patients. This study will investigate whether warming the hand and lower part of the forearm can improve the accuracy of pulse oximeter measurements in mechanically ventilated adult intensive care patients with low blood flow to the fingers. Participants will be randomly assigned to either a local warming strategy or usual care without warming. In the warming group, the hand and lower forearm will be warmed for 15 minutes to a target temperature of 41-42°C. Pulse oximeter measurements will be compared with oxygen saturation measured from an arterial blood sample before and after the intervention. The main objective is to determine whether local warming reduces the difference between pulse oximeter oxygen saturation (SpO₂) and arterial oxygen saturation (SaO₂) 15 minutes after the start of the assigned strategy. The study will also assess whether any improvement persists at 60 minutes, how peripheral blood flow changes during 4 hours of follow-up, whether repeat warming is effective when blood flow decreases again, and whether the effect differs according to skin phototype. Safety related to the warming intervention will also be evaluated.

Pulse oximetry is a fundamental component of monitoring and oxygen titration in critically ill patients. Its accuracy may, however, be impaired by reduced peripheral perfusion. Low peripheral perfusion is common in mechanically ventilated intensive care patients and may contribute to clinically relevant differences between peripheral oxygen saturation measured by pulse oximetry (SpO₂) and arterial oxygen saturation measured by blood gas analysis (SaO₂). Local warming increases cutaneous blood flow and is routinely used to improve peripheral perfusion during procedures such as capillary blood sampling. Preliminary single-center data suggest that local warming may also improve pulse oximetry accuracy in critically ill patients with low peripheral perfusion. Whether this effect is present when evaluated in a randomized controlled trial is unknown. This is a prospective, multicenter, parallel-group randomized controlled strategy trial conducted in four intensive care units. Mechanically ventilated adult ICU patients with an arterial catheter and a finger perfusion index (PFI) \<1.0 will be randomized in a 1:1 ratio to a protocolized local warming strategy or usual care without protocolized warming. Randomization will be stratified by study site and Fitzpatrick skin phototype. In the intervention group, local warming will be applied to the dorsum of the hand and distal forearm for 15 minutes, with a target interface temperature of 41-42°C. Repeat warming is permitted during follow-up if finger PFI decreases below 0.5 according to the prespecified study strategy. The control group will receive usual care without protocolized local warming. SpO₂, finger PFI, earlobe PFI, and arterial SaO₂ will be assessed at baseline, 15 minutes, and 60 minutes. Additional PFI measurements will be obtained at 120 and 240 minutes. Oxygen therapy will otherwise be managed according to standard clinical care. The primary objective is to compare pulse oximetry bias between the randomized groups at 15 minutes. Bias is defined as SpO₂ minus SaO₂, expressed in percentage points. The primary analysis will compare the groups using an analysis of covariance adjusted for baseline SpO₂-SaO₂ bias and study site. Secondary objectives include assessment of pulse oximetry accuracy at 60 minutes, the proportion of measurements within ±2 percentage points of SaO₂, changes in peripheral perfusion during 4 hours of follow-up, the local versus systemic perfusion response assessed using finger and earlobe PFI, the response to repeat warming, safety outcomes, and possible treatment-effect modification according to Fitzpatrick skin phototype. The planned sample size is 240 participants, with 120 participants allocated to each study group.