Article Text
Abstract
Aim To describe how the stability of oxygen saturation measured by pulse oximetry (SpO2%) varies within and between infants with bronchopulmonary dysplasia (BPD).
Methods Clinically stable infants with BPD had SpO2 measured at different inspired oxygen concentrations (FIO2 expressed as %). A computer model of gas exchange, that is, ventilation/perfusion ratio (VA/Q) and shunt, plotted the curve of SpO2 versus FIO2 best fitting these data. The slope of this curve is the change in SpO2 per % change in FIO2, hence SpO2 stability, calculated at each SpO2 from 85% to 95%.
Results Data from 16 infants with BPD previously described were analysed. The dominant gas exchange impairment was low VA/Q (median 0.35, IQR, 0.16–0.4, normal 0.86). Median shunt was 1% (IQR, 0–10.5; normal <2%). Slope varied markedly between infants, but above 95% SpO2 was always <1.5. In infants with least severe BPD (VA/Q ≈0.4, shunt ≤2%) median slope at 85% SpO2 was 5.1 (IQR, 3.7–5.5). With more severe BPD (VA/Q ≤0.3) slope was flatter throughout the SpO2 range. The highest FIO2 for 90% SpO2 was in infants with the lowest VA/Q values.
Conclusions In infants with BPD, there was large variation in the slope of the curve relating SpO2% to inspired oxygen fraction in the SpO2 range 85%–95%. Slopes were considerably steeper at lower than higher SpO2, especially in infants with least severe BPD, meaning that higher SpO2 target values are intrinsically much more stable. Steep slopes below 90% SpO2 may explain why some infants appear dependent on remarkably low oxygen flows.
- Fetal Medicine
- Intensive Care
- Monitoring
- Physiology
Statistics from Altmetric.com
What is already known about this topic
Targeting SpO2 in oxygen-dependent preterm infants is difficult.
In clinical trials, infants targeted to lower SpO2 spent less time in their intended SpO2 target range than infants targeted to higher SpO2.
Data series of FIO2 versus SpO2 can be used to calculate shunt and the ratio of ventilation to perfusion (VA/Q).
What this study adds
The slope of the relationship between inspired oxygen pressure (FIO2) and SpO2 is highly variable between infants with bronchopulmonary dysplasia.
In infants with least severe disease (VA/Q≈0.4), small changes in FIO2 predispose to much larger change in SpO2 at 85% than 95% SpO2.
Higher SpO2 target values are likely to be intrinsically more stable and easier to achieve than lower SpO2 values.
Introduction
Preterm infants with bronchopulmonary dysplasia (BPD) remain oxygen dependent for long periods. Adjusting the inspired oxygen concentration (FIO2) to maintain pulse oximeter saturation (SpO2) within a desired range is difficult, and SpO2 in some infants is much less stable than others.1 Low and high SpO2 have been linked to mortality and morbidity.2–6 Consequently, it is of clinical importance to understand what determines the stability of the SpO2 at any given saturation as this affects the set target value and the method used to supplement FIO2. Such instability also helps explain why some infants are difficult to wean from oxygen because they desaturate even when minimal supplemental oxygen flows are discontinued.
We have previously shown in infants with BPD that a series of paired measurements of SpO2 and FIO2 can be plotted as a curve and used to calculate the ratio of alveolar ventilation relative to perfusion (VA/Q) and shunt.7 This model, derived mathematically from basic physiology, was first described diagrammatically and then computerised with a graphical output.8–12 The method has been validated against more invasive measures of gas exchange.13–16 Since the earlier report, our methodology has been extended to describe less homogeneous lung disease with regions of lung with different degrees of reduced VA/Q.17 ,18 The updated system has been used to measure gas exchange impairment in a further population of infants with BPD.19
The normal curve of SpO2 against FIO2 takes its shape from the oxyhaemoglobin dissociation curve. A uniform reduction in VA/Q from 0.86 to 0.4 shifts the whole curve from A to B (figure 1). We have shown that a very unstable SpO2 results from a similar reduction in VA/Q in adults breathing air.9 ,20 This instability is explained because the curve, without change in shape, has shifted to the right, and at the point where FIO2 is 21%, curve B is 10 times steeper than curve A. In contrast, non-homogeneous changes in VA/Q result in hybrid curves interpreted by a computer analysis as one or two homogeneous lung compartments with a shunt.17 At 90% SaO2, curves D and E are much flatter than B.
The slope at any point on the curve of SpO2 against FIO2 is the change in SpO2 that will occur for a 1% change in FIO2 at that point. We use the word ‘slope’ exclusively to refer to the slope of this curve and, for simplicity, report it as a dimensionless quantity. Understanding how this slope varies between infants and at different target SpO2 values may explain much of the variation in stability of SpO2. Changes in breathing pattern, lung volume or pathology, which result in desaturation, do so by reducing VA/Q and/or increasing intrapulmonary shunt. The resulting alveolar hypoxia may have further adverse effects on the pulmonary vasculature. VA/Q and shunt determine the FIO2 required to achieve a normal alveolar oxygen pressure and overcome desaturation.
We calculated the slope at each SpO2 in the range 85%–95% for each infant with BPD and demonstrate how it varied in this SpO2 range within and between infants with different VA/Q and shunt.
Methods
We reanalysed the data gathered from preterm infants with BPD in our previous report using the new fully computerised system.12 ,17 The infants were studied between September 2006 and December 2007 in the neonatal unit of the Simpson Centre for Reproductive Health, Royal Infirmary of Edinburgh. Ten subjects were male and six were female. Their mean (range) gestational age at birth was 26+4 weeks (23+4–29+4) and birth weight was 929 g (510–1335 g). Data series of FIO2 versus SpO2 pairs were obtained within 3 days of reaching 36 weeks postmenstrual age. The infants were dependent on supplemental oxygen at the time of study. With the infant lying quietly supine in an incubator, FIO2 was reduced in steps and the representative SpO2 observed in a steady state at each FIO2 was determined as previously described.7 The study was approved by an ethics advisory committee and written informed consent was obtained from the parents. For each infant, we plotted the SpO2 versus FIO2 curve and calculated the slope at each SpO2 in the range 85%–95%. We calculated the percentage shunt, VA/Q, and for infants with curves indicating uneven lung pathology, the fraction of cardiac output perfusing the low VA/Q compartment.
Statistical comparisons were made using the R software package, V.3.2.2.
Results
Infants were ordered by the value of slope at 85% SpO2 in the first row (table 1). In all infants, the slope was <1.5 at a SpO2 of 95%, but the variation in slope at 90% SpO2 was 17-fold (0.2–3.4). VA/Q correlated strongly and inversely with the FIO2 required to achieve 90% SpO2 (Pearson, p<0.001). VA/Q also correlated with the slope at 90% SpO2 (p<0.005), but not with shunt. To be explicit, infants with smaller reduction of VA/Q (less disease) had the lowest FIO2 needed to achieve 90% SpO2 but also had the steepest slopes. Infants with the greatest reduction in VA/Q (worse disease) needed the most FIO2 to achieve 90% SpO2, but had the flattest slopes.
Infants 6, 9, 11, 12, 13, 15 and 16 satisfied the NIH criterion for severe BPD21 (≥30% FIO2 required to achieve 90% SpO2); their VA/Q values and slopes at 90% SpO2 were significantly less than the others (Kruskal–Wallis, p<0.01), though there was no difference in shunt.
In infants 1–11, the two-compartment and three-compartment computer models produced the same curve for the FIO2 versus SpO2 data points, implying homogeneous lung disease. These infants had larger VA/Q and slopes at 90% SpO2 than those requiring three-compartment analysis (Kruskal–Wallis, p<0.05); the latter had slopes <1 throughout the SpO2 range 85%–95%. There was no difference in shunt or FIO2 required to achieve 90% SpO2.
Figure 2 shows the plots of slope on the vertical axis against oxygen saturation on the horizontal axis. Infants with the flattest slopes, <1 at 90% SpO2, have the worst gas exchange.
Figure 3 shows individual plots of SpO2 versus FIO2. Infant 4 with only a small reduction in VA/Q had a steep curve at 90% SpO2 predisposing to a very unstable SpO2 breathing air. Infant 8 had a similar VA/Q but a 20% shunt flattened the curve above 90% SpO2 effectively stabilising the saturation at that level. Infant 11 had a considerable reduction in VA/Q and needed high FIO2 (>58%) to exceed 90% SpO2. A 20% shunt flattened the curve above 90% SpO2, but the curve below 90% SpO2 became very steep. The three-compartment model fit for infant 13 is shown by the heavy line in figure 3.
Discussion
The slope of the relationship between SpO2 and inspired oxygen pressure was highly variable between these infants with BPD because of differences in VA/Q and shunt. This is an important determinant of the way that a change in breathing pattern will influence SpO2 and of the likely response to a change in inspired oxygen pressure.
Some infants with milder disease had steeply sloping curves within the range of clinically targeted SpO2 values and this may explain why they can be stable with minimal flow oxygen but desaturate markedly when it is discontinued. In most infants, the slope was considerably greater at lower SpO2. At a SpO2 of 85%, a 1 kPa change in FIO2 could change SpO2 in an infant with BPD by >5%, whereas at 95% SpO2, the same change in FIO2 would commonly change SpO2 by <1%. This means that higher target SpO2 values are intrinsically more stable than lower values, and may partly explain why, in the recent randomised trials of oxygen targets, infants in the lower saturation target groups spent less time in their intended target range than infants targeted to higher SpO2.3 The finding that the sicker infants will tend to be less sensitive to small changes in inspired oxygen than those with milder lung disease is counterintuitive, but it is potentially important in the planning of clinical management strategies for oxygen adjustment and in the interpretation of studies of SpO2. It may help to inform the design of future automated oxygen adjustment systems.
All of the infants in this report had BPD where the dominant gas exchange abnormality is a reduced VA/Q.7 ,12 ,13 ,19 Earlier in the clinical course, when the alveoli are more unstable and there is more intrapulmonary shunting, the same method could be applied but the slopes would be predicted to be flatter. It would be valuable to use this method longitudinally to show how the slope changes through the clinical course.
The absence of shunt in the five infants with the lowest VA/Q and requiring three-compartment analysis is interesting. Their median FIO2 to maintain 90% SpO2 was 41%, which is functionally equivalent to a >30% shunt at SpO2 90% (see curve E, figure 1B). Using the nitrogen gradient method in preterm infants to measure VA/Q directly, Hand et al22 showed that VA/Q ratios could be as low as 0.05 and may account for almost all of the oxygen gradient. Our five infants had a very low VA/Q (median 0.09) perfused by nearly 50% pulmonary blood flow. The three-compartment model can resolve a 20% shunt when 50% non-shunt flow passes through a VA/Q of 0.2. However, when VA/Q is very low (≤0.16), the three-compartment model may not be able to differentiate between shunt and VA/Q.
In conclusion, we have shown that in oxygen-dependent preterm infants with BPD, the slope of the relationship between FIO2 and SpO2 is highly variable between infants. The slope is usually much steeper at lower SpO2 values than at higher values, and consequently, lower SpO2 values are likely to be more unstable and more difficult to target. In infants with least severe BPD (VA/Q ≈ 0.4), the slope can be particularly steep at values below 90%, and this helps to explain why convalescent preterm infants can be dependent on remarkably small amounts of supplemental oxygen.
Acknowledgments
We thank Professor Colin Morley for help in the preparation of this paper. The updated software package can be obtained on request from GGL (g.lockwood@imperial.ac.uk).
References
Footnotes
Contributors JGJ: conceived the study, analysed the data and wrote paper. GGL: developed the computer method and analysed the data. NF and JL: developed the computer method. RIRR: conceived the method and revised the paper. DQ: collected data and revised the paper. BJS: conceived the study and wrote the paper.
Competing interests None declared.
Patient consent Obtained.
Ethics approval Lothian Local Research Ethics Committee, Royal Infirmary of Edinburgh NHS Trust.
Provenance and peer review Not commissioned; externally peer reviewed.