Prediction of PaO2 from SpO2 values in critically ill invasively ventilated patients: rationale and protocol for a patient–level analysis of ERICC, LUNG SAFE, PRoVENT and PRoVENT–iMiC (PRoPERLy II)

TL;DR

This study aims to compare methods for predicting arterial oxygen levels using pulse oximetry data in critically ill patients.

Contribution

The novelty lies in evaluating the accuracy of PaO2 prediction equations from SpO2 in a pooled database of four critical care studies.

Findings

The study will assess the accuracy of existing equations to predict PaO2 from SpO2 in critically ill patients.

It will compare risk stratification using measured versus predicted PaO2/FiO2 ratios.

Findings could influence clinical practices in settings with limited access to arterial blood gas analysis.

Abstract

In patients with acute respiratory distress syndrome (ARDS), mortality risk is typically assessed using the arterial partial pressure of oxygen (PaO2) divided by the fraction of inspired oxygen (FiO2), known as the PaO2/FiO2 ratio. Recently, the SpO2/FiO2 ratio, which uses peripheral oxygen saturation (SpO2) instead of PaO2, has been suggested as a reasonable alternative when there is little access to arterial blood gas analyses. Additionally, equations that predict PaO2 from SpO2 values could offer another viable method for assessment.

To evaluate the accuracy of methods for predicting PaO2 from SpO2 values and compare risk stratification based on measured versus predicted PaO2/FiO2 ratios using a large database that harmonizes the individual data of patients included in four observational studies.

The individual data from four observational studies (‘Epidemiology of Respiratory Insufficiency in Critical Care study’ [ERICC], ‘Large Observational Study to Understand the Global Impact of Severe Acute Respiratory Failure’ [LUNG SAFE], ‘PRactice of VENTilation in critically ill patients without ARDS’ [PRoVENT], and ‘PRactice of VENTilation in critically ill patients in Middle–income Countries’ [PRoVENT–iMiC]) were harmonized and pooled into a database named ‘PRoPERLy II’. The primary endpoint of this planned analysis will be the accuracy of currently available methods to predict PaO2 from SpO2 values. The secondary endpoint will be the accuracy of classification based on SpO2/FiO2 ratio and the predicted PaO2/FiO2 ratio.

Our planned analysis addresses a clinically important research question by comparing different equations for predicting PaO2 from SpO2 values.

Figures