Article Text


Individualised pulse oximetry limits in neonatal intensive care
  1. R Gupta,
  2. C W Yoxall,
  3. N Subhedar,
  4. N J Shaw
  1. Neonatal Intensive Care Unit Liverpool Women’s Hospital Liverpool L8 7SS
  1. Dr N J Shaw.


AIM To determine whether individualised limits for arterial oxyhaemaglobin saturation by pulse oximetry (SpO2) are more effective for detecting hypoxia and hyperoxia in sick newborn infants than setting fixed limits.

METHODS Six hundred and ninety two simultaneous measurements of SpO2 and partial pressure of oxygen in arterial blood (PaO2) were made in 95 infants. The sensitivity and specificity for predicting hypoxia and hyperoxia using various fixed SpO2 limits and also individualised SpO2 limits, calculated using a standard equation, were determined and compared.

RESULTS None of the fixed limits for SpO2 was both sensitive and specific for predicting hypoxia and/or hyperoxia. There was no difference between these and individualised limits.

CONCLUSION Individualised SpO2 limits are no more effective than fixed SpO2 limits for predicting hypoxia and/or hyperoxia in sick newborn infants. SpO2 monitoring is not an ideal method for assessing PaO2.

Statistics from

The partial pressure of oxygen in the arterial blood (PaO2) is the gold standard measurement of oxygenation in sick babies. However, pulse oximetry, to prevent chronic hypoxia and hyperoxia, is widely used to monitor arterial oxygenation in neonatal intensive care units.1 2 The position of the oxygen–haemoglobin dissociation curve (OHDC) is influenced by several factors—for example, temperature, pH, percentage haemoglobin F and partial pressure of carbon dioxide in the arterial blood (PaCO2).3 The curve may therefore alter during the course of an infant’s illness, making the estimation of arterial oxyhaemoglobin saturation using pulse oximetry (SpO2) a poor predictor of PaO2. This, together with the sigmoid shape of the OHDC, makes it difficult to set acceptable upper and lower limits for SpO2. If the shape of an individual infant’s OHDC could be determined by the most recent arterial blood gas result, thus taking into account the aforementioned influential factors, it might be possible to “customise” limits for an individual baby at any time by measuring SpO2 and PaO2simultaneously, using a standard equation for the shape of the OHDC to predict which SpO2 levels correspond to the desired PaO2 range.

This study aimed to determine whether such individualised limits are more effective at detecting hypoxia and hyperoxia in sick babies than setting fixed limits.


Infants admitted to the neonatal intensive care units at the Liverpool Maternity Hospital and the Liverpool Women’s Hospital, and who required arterial lines, were eligible for the study. Over three years (July 1994 to August 1997) simultaneous measurements of SpO2 and PaO2 were prospectively made by recording the infant’s name and the SpO2 reading (from the Ohmeda oximeter being used on the infant for clinical purposes) on the blood gas result printout, when an arterial blood sample was taken for blood gas analysis. The SpO2 reading had to be stable for at least 30 seconds before the arterial blood sample could be taken, thus making it highly unlikely that the infant’s SpO2would be on the steep part of the OHDC (suggesting clinically significant hypoxia) at the time of sampling. As this was a pragmatic study, the results of which would hopefully be clinically applicable, no infant with hypotension was excluded. However, infants with structural congenital cardiac lesions were not studied. The date and time of blood sample analysis was automatically printed by the blood gas analyser. All blood gas printouts were stored for later analysis.

For each baby at least two sets of measurements were required, taken a maximum of 6 hours apart. For each pair of measurements the sensitivity and specificity of fixed SpO2 limits for predicting hypoxia and hyperoxia were calculated. Individualised SpO2 limits were then calculated retrospectively using a standard equation for the shape of the OHDC, devised by Tien and Gabel.4 These were then applied to the next pair of measurements in that individual. The sensitivity and specificity of the individualised limits for detecting hypoxia and hyperoxia were calculated and compared with those when using fixed limits.

Hypoxia was defined as a PaO2 of less than 6 kPa and hyperoxia as a PaO2 of more than 10 kPa, using the recommendations of the Joint Working Group of the British Association of Perinatal Medicine and the Research Unit of the Royal College of Physicians.5


During the study period 692 simultaneous measurements of SpO2 and PaO2 were made in 95 infants. The median (range) gestation was 28 (23 to 40) weeks and the median (range) birthweight was 1165 (590 to 3570) g. Thirty girls and 65 boys were studied. The sensitivities and specificities for hypoxia and hyperoxia for fixed and individualised limits for SpO2 are shown in tables 1 and 2.

Table 1

Hypoxia prediction

Table 2

Hyperoxia prediction


To prevent potential morbidity in preterm infants, associated with episodes of chronic hypoxia and/or hyperoxia while ventilated, we studied the effect of using individualised SpO2 limits, but were unable to show any benefit compared with the use of fixed SpO2 limits.

Theoretically, because of the variable relation between PaO2 and arterial oxyhaemoglobin saturation (SaO2) in sick infants, recalculating the shape of the OHDC in each infant after a blood gas analysis result should take into account some of the factors which alter its shape, and therefore provide SpO2 limits which are more appropriate for that infant. This should result in a more effective method to prevent hypoxia and hyperoxia in these patients. Unfortunately, in clinical practice, setting individualised limits, using the method described, was no more effective at predicting hypoxia than a lower fixed limit of 91%, and it was no better at predicting hyperoxia than an upper fixed limit of 96%.

There are several possible reasons for this. Setting individualised limits requires accurate simultaneous measurements of both PaO2 and SpO2. In clinical practice, on a busy neonatal unit, some of the measurements will probably not be accurate. Imprecisions in PaO2 measurements using standard blood gas analysers exist,6 and these may be greater on a busy neonatal unit when techniques for sampling and processing may not always be optimal. The SpO2 values produced correspond to real SaO2 values within about ± 2% and pulse oximetry is notoriously prone to artefact.7 Furthermore, the equation used to describe the shape of the OHDC may be inaccurate. However, of three equations for calculating PaO2 from SaO2, assessed by Gabel, the equation produced by Tien and Gabel was found to be most accurate.8 Finally, the position of the OHDC may move frequently in an individual infant. Some of the infants in the study would have received red cell transfusions between blood gas measurements and this may have affected the position of the OHDC by altering the amount of circulating haemoglobin F. Although this theoretically affects the OHDC, it has not been shown to have a major effect clinically,9 and is unlikely to have affected our results.

On a practical level, individualising SpO2 limits has its difficulties. Firstly, calculating individualised limits is complex. Secondly, the calculated limits would have to be communicated to the member of nursing staff looking after the infant on each occasion, and care would have to be taken to ensure that the limits were communicated at each nursing and medical handover. Thirdly, continuously changing limits may lead to confusion.

Individualised SpO2 limits could, however, be used on an ad hoc basis if there is a trend towards too high or too low PaO2 readings compared with the monitored SpO2readings in a particular infant. This should be reviewed frequently especially if the clinical state of the infant changes.

Oxygen saturation monitors were not designed specifically for preventing hypoxia and hyperoxia. A better way of non-invasively assessing PaO2 than saturation monitoring is transcutaneous oxygen monitoring. However, many units use saturation monitors to monitor oxygenation and the sensitivities and specificities in tables 1and 2 can be used to help determine the most appropriate limits for SpO2. Our results suggest that the best fixed lower and upper limits for SpO2 are 91% (sensitivity of 0.72 and specificity of 0.56) and 96% (sensitivity of 0.80 and specificity of 0.59), respectively, but these should be used only as a guide. Previous authors have suggested lower limits for SpO2 ranging from 89 to 93%,3 10 and upper limits for SpO2ranging from 95 to 97%.3-11

In summary, when using SpO2 monitoring to assess oxygenation, individualised SpO2 limits are no better than fixed limits for preventing hypoxia or hyperoxia.


View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.