Peripheral oxyhemoglobin saturation (SpO₂) instability may result in more than 100 hypoxemic events per day within the first 8 weeks of life in preterm infants (Di Fiore et al., 2010). Moreover, SpO₂ instability is poorly documented despite the correlation with an increased rate of late death or disability at 18 months of age (Poets et al., 2015). SpO₂ instability increases when the infant is in supine compared to the prone position, independent of the mode of respiratory support (Miller-Barmak et al., 2020). A contributing factor for hypoxemic episodes in preterm infants includes sudden decrease in lung volume leading to small airway collapse and intrapulmonary shunt (Bolivar et al., 1995). We have previously shown in adults that SpO₂ was unstable when ventilation to perfusion ratio (V_A/Q) was reduced, but was stable with increased shunt (Jones and Jones, 2000).

The concept of V_A/Q is well established but infrequently used in neonatal practice, as the technics for measuring V_A/Q in infants are technically difficult e.g. A-a nitrogen difference, or impossible e.g. Multiple Inert Gas Elimination Technique (MIGET) (Corbet et al., 1974; Hand et al., 1990; Roca and Wagner, 1994). Our non-invasive method for measuring V_A/Q in adults is based on the different effects on SpO₂ of changing inspired oxygen pressure (P _IO₂) when either V_A/Q is reduced or shunt increased (Jones and Jones, 2000). The method depends on the shape of the oxyhaemoglobin dissociation curve (ODC). We adapted this method for preterm infants by using the neonatal rather than the adult ODC as the reference and plotting SpO₂ at different P _IO₂ (Lockwood et al., 2014). A computer algorithm derived a model of V_A/Q and shunt from the paired SpO₂ and P _IO₂ dataset. Decreasing V_A/Q is reflected by a right shift of the SpO₂ vs. P _IO₂ curve. The right shift leads to a steeper slope of the curve as it intersects the 21 kPa P _IO₂ line (P _IO₂ = F_IO₂ x (barometric pressure—saturated water vapour pressure), which means 21 kPa P _IO₂ is equal to room air at sea level (Figure 1). Any paired SpO₂ vs. P _IO₂ measurement located on the steep section of the ODC may predispose the infant to SpO₂ instability with small changes in alveolar oxygen.

We hypothesised that SpO₂ instability in preterm infants with cardiorespiratory disease is not an epiphenomenon but an important clinical sign of a reduced V_A/Q rather than right to left intrapulmonary shunt. We used a new algorithm and routine SpO₂ monitoring to derive V_A/Q and shunt.

We conducted a prospective observational study in two phases. Firstly, we used a mathematical model of pulmonary gas exchange to examine changes in V_A/Q or increasing shunt on the slope of the SpO₂ vs. P _IO₂ curve.

Secondly, we recorded 24 hourly SpO₂ vs. P _IO₂ measurements in extremely preterm infants requiring continuous SpO₂ monitoring. From these measurements, we derived V_A/Q and shunt using a pulmonary gas exchange algorithm (Lockwood et al., 2014).

We suggest that a reduced V_A/Q predisposes preterm infants to SpO₂ instability. The pulmonary gas exchange model showed that reducing V_A/Q shifted the P _IO₂ vs. SaO₂ curve to the right, whereas shunt displaced the curve downwards. Right shift increased the effective curve slope up to 16 times where it intercepted the 21 kPa P _IO₂ line e.g., breathing room air. The right shift predicted large fluctuations in SaO₂ with small changes in alveolar oxygen. Reducing V_A/Q from 0.4 to 0.3 predicted a 24% fall in SaO₂ breathing air. In contrast a large shunt had little effect on curve slope and a small fall in SaO₂.

These predictions were confirmed in extremely preterm infants by routine monitoring of SpO₂ and P _IO₂. The paired SpO₂ vs. P _IO₂ measurements were used to derive V_A/Q, right shift and shunt using Lockwood’s algorithm (Lockwood et al., 2014). All infants had a reduced V_A/Q. The SpO₂ was unstable with a median (IQR) 8.0 (7.0) % at a constant P _IO₂ compared to a mean (±SD) of 3% (±1.5%) in healthy newborn infants breathing room air (Dassios et al., 2017; Dassios et al., 2019). Infants with V_A/Q < 0.52 had considerable increase in SpO₂ variability, seven infants had episodes with SpO₂ ≤ 84%, whilst four of these infants were dependent on supplemental oxygen ≥27%. Two infants had transient shunt lasting for a few hours, which was not associated with SpO₂ instability. In the remaining infants, the P _IO₂ was insufficiently high to fully characterize shunt due to the SpO₂ target range (90%–94%).

The non-invasive method for deriving V_A/Q and shunt normally depends on varying P _IO₂ (Rowe et al., 2010; Bamat et al., 2015; Dassios et al., 2017; Dassios et al., 2019; Svedenkrans et al., 2019; Bamat et al., 2022). More recently, we showed that V_A/Q and/or shift can be derived from serial SpO₂ measurements at a fixed P _IO₂ (Svedenkrans et al., 2019; Stoecklin et al., 2021). A reduced V_A/Q lowers SpO₂ at any given P _IO₂ and increases the effective slope of the SpO₂ vs. P _IO₂ curve, but does not by itself make SpO₂ unstable unless V_A/Q itself is unstable (Jones, 2021). The reduced V_A/Q amplifies the destabilizing effects on SpO₂ during changes in cardiac output, ventilation, central or obstructive apnea and posture. Interestingly, the 10 infants with the most unstable SpO₂ were equally distributed between those breathing spontaneously and those on mechanical ventilation. Reduced V_A/Q can lead to SpO₂ instability independent of the mode of respiratory support used in an infant.

In agreement with our results, a reduced V_A/Q in adults leads to SpO₂ instability breathing room air. By adding a time dimension, the different effects of V_A/Q or shunt in individual adult patients breathing air is more clearly illustrated by dynamic waterfall plots of continuously monitored SpO₂ (Entwistle et al., 1991; Jones and Jones, 2000; Jones, 2021) than do “static” histograms (Borenstein-Levin et al., 2020). A typical example of unstable SpO2 in a patient with reduced V_A/Q breathing air is shown in Figure 3 in Ref (Jones, 2021). This patient had blunt curves with SpO₂ fluctuating from 77% to 95%. In contrast, a patient with increased shunt had superimposed stable SpO₂ peaks within a much narrower SpO₂ range.

We note that the right to left shunt in our study might include some element of cardiac shunting occurring via an open ductus arteriosus. Our method does not allow differentiation between intrapulmonary or cardiac shunt. However, in the majority of preterm infants the ductus arteriosus would have functionally closed on day seven of life. In the few infants with a persistent ductus arteriosus, at day seven of life a predominantly left to right shunt would be expected which would not affect our calculations (de Klerk et al., 2020).

The clinical applicability of our findings is that routine measurements of SpO₂ and P _IO₂ can be analysed with our algorithm to derive V_A/Q and shunt, thereby characterizing respiratory disease in preterm infants. V_A/Q and shunt incorporate information on the mechanisms of hypoxemia and explain SpO₂ instability. Consequently, SpO2 instability can be prevented by a small increase in P _IO₂.

In conclusion, we reported that predisposition to oxygen saturation instability in preterm infants results from a reduced ventilation to perfusion ratio rather than from increased intrapulmonary shunt. We have highlighted how routine monitoring can be used to derive non-invasive measurements of oxygenation impairment in preterm infants. Future research should aim at strategies to not only detect SpO₂ instability, but to also prevent such instability.