Article Text
QUESTION: Are individualised limits for arterial oxyhaemaglobin saturation (SpO_{2}) assessed by pulse oximetry as accurate as fixed limits for detecting hypoxia and hyperoxia in sick newborn infants?
Design
Comparison of sensitivities and specificities of individualised and fixed SpO_{2} limits for detecting hypoxia or hyperoxia.
Setting
A neonatal intensive care unit in Liverpool, UK.
Patients
95 infants (68% boys, median gestational age 28 weeks, median birth weight 1165 g) who required arterial lines. Infants with structural congenital cardiac lesions were excluded.
Description of tests and diagnostic standard
Simultaneous measurements of SpO_{2} and partial pressure of oxygen in arterial blood (PaO_{2}) (diagnostic standard) taken over a 3 year period. For each infant, ≥2 sets of measurements were required, taken ≤6 hours apart. For each pair of measurements, sensitivity and specificity of fixed SpO_{2} limits for detecting hypoxia (PaO_{2} <6 kPa) and hyperoxia (PaO_{2} >10 kPa) were calculated. Individualised SpO_{2} limits were calculated retrospectively using a standard equation for the shape of the oxygenhaemoglobin dissociation curve; these limits were then applied to the next pair of measurements in that infant.
Main outcome measures
Sensitivity and specificity of individualised and fixed limits of SpO_{2} for detecting hypoxia and hyperoxia.
Main results
A total of 692 simultaneous measurements of SpO_{2} and PaO_{2} were taken. Individualised SpO_{2} lower limits for hypoxia had similar test properties (sensitivities and specificities) to a fixed lower limit of 91% and individualised upper limits for hyperoxia were similar to a fixed upper limit of 96% (table).
Conclusion
When monitoring arterial oxyhaemaglobin saturation by pulse oximetry, individualised limits were no better than fixed limits for detecting hypoxia or hyperoxia in sick infants.
Commentary
In premature and sick newborn infants, hypoxia is associated with organ damage and pulmonary hypertension and hyperoxia is associated with retinopathy of prematurity.^{1} It is important, therefore, to have an accurate measure for monitoring arterial oxygen levels so that appropriate levels can be maintained and adverse events prevented. Pulse oximetry is the standard for monitoring SpO_{2} in newborns.^{2} Gupta et al examined whether fixed SpO_{2} limits for identifying hypoxia and hyperoxia were as good as individualised limits derived from previous readings for a given infant. The derivation of these individualised limits was based on a complex formula that accounted for the oxygenhaemoglobin curve.
The study results indicate that the sensitivities and specificities of individualised limits were similar to those of fixed limits. Sensitivity and specificity values can, however, be misleading when used as a basis for clinical decision making because they restrict us to only 2 levels of results (positive and negative).^{3} In this study, many different levels (limits) were assessed, and therefore, it may be more informative to consider likelihood ratios (LRs), which tell us how much the probability of hypoxia (or hyperoxia) changes from baseline when the test is positive (positive LR) or negative (negative LR). Positive LRs >2 and negative LRs <0.1 should be considered in clinical decision making. As we can see from the LRs calculated from the data in this study, neither fixed nor individualised lower limits for detecting hypoxia had positive LRs >2 or negative LRs <0.1, and fixed and individualised limits had similar LRs for SpO_{2} limits of 89% (positive LR) and 90% (negative LR). For hyperoxia, a fixed upper limit of 97% and individualised limits both had positive LRs >2, whereas fixed limits of 94% and 95% had negative LRs <0.1. Given this, it does not appear that that the additional work required to establish individualised SpO_{2} limits resulted in better accuracy for identification of hypoxia or hyperoxia.
Nevertheless, the results of this study will be useful to nurses working in NICUs in determining optimal upper and lower limits for setting alarms. When determining these limits, clinicians should also be aware that hypothermia, cardiac anomalies, partial pressure of carbon dioxide, pH, and haemoglobin levels can affect the accuracy of pulse oximetry.^{4}
Statistics from Altmetric.com
QUESTION: Are individualised limits for arterial oxyhaemaglobin saturation (SpO_{2}) assessed by pulse oximetry as accurate as fixed limits for detecting hypoxia and hyperoxia in sick newborn infants?
Design
Comparison of sensitivities and specificities of individualised and fixed SpO_{2} limits for detecting hypoxia or hyperoxia.
Setting
A neonatal intensive care unit in Liverpool, UK.
Patients
95 infants (68% boys, median gestational age 28 weeks, median birth weight 1165 g) who required arterial lines. Infants with structural congenital cardiac lesions were excluded.
Description of tests and diagnostic standard
Simultaneous measurements of SpO_{2} and partial pressure of oxygen in arterial blood (PaO_{2}) (diagnostic standard) taken over a 3 year period. For each infant, ≥2 sets of measurements were required, taken ≤6 hours apart. For each pair of measurements, sensitivity and specificity of fixed SpO_{2} limits for detecting hypoxia (PaO_{2} <6 kPa) and hyperoxia (PaO_{2} >10 kPa) were calculated. Individualised SpO_{2} limits were calculated retrospectively using a standard equation for the shape of the oxygenhaemoglobin dissociation curve; these limits were then applied to the next pair of measurements in that infant.
Main outcome measures
Sensitivity and specificity of individualised and fixed limits of SpO_{2} for detecting hypoxia and hyperoxia.
Main results
A total of 692 simultaneous measurements of SpO_{2} and PaO_{2} were taken. Individualised SpO_{2} lower limits for hypoxia had similar test properties (sensitivities and specificities) to a fixed lower limit of 91% and individualised upper limits for hyperoxia were similar to a fixed upper limit of 96% (table).
Conclusion
When monitoring arterial oxyhaemaglobin saturation by pulse oximetry, individualised limits were no better than fixed limits for detecting hypoxia or hyperoxia in sick infants.
Commentary
In premature and sick newborn infants, hypoxia is associated with organ damage and pulmonary hypertension and hyperoxia is associated with retinopathy of prematurity.^{1} It is important, therefore, to have an accurate measure for monitoring arterial oxygen levels so that appropriate levels can be maintained and adverse events prevented. Pulse oximetry is the standard for monitoring SpO_{2} in newborns.^{2} Gupta et al examined whether fixed SpO_{2} limits for identifying hypoxia and hyperoxia were as good as individualised limits derived from previous readings for a given infant. The derivation of these individualised limits was based on a complex formula that accounted for the oxygenhaemoglobin curve.
The study results indicate that the sensitivities and specificities of individualised limits were similar to those of fixed limits. Sensitivity and specificity values can, however, be misleading when used as a basis for clinical decision making because they restrict us to only 2 levels of results (positive and negative).^{3} In this study, many different levels (limits) were assessed, and therefore, it may be more informative to consider likelihood ratios (LRs), which tell us how much the probability of hypoxia (or hyperoxia) changes from baseline when the test is positive (positive LR) or negative (negative LR). Positive LRs >2 and negative LRs <0.1 should be considered in clinical decision making. As we can see from the LRs calculated from the data in this study, neither fixed nor individualised lower limits for detecting hypoxia had positive LRs >2 or negative LRs <0.1, and fixed and individualised limits had similar LRs for SpO_{2} limits of 89% (positive LR) and 90% (negative LR). For hyperoxia, a fixed upper limit of 97% and individualised limits both had positive LRs >2, whereas fixed limits of 94% and 95% had negative LRs <0.1. Given this, it does not appear that that the additional work required to establish individualised SpO_{2} limits resulted in better accuracy for identification of hypoxia or hyperoxia.
Nevertheless, the results of this study will be useful to nurses working in NICUs in determining optimal upper and lower limits for setting alarms. When determining these limits, clinicians should also be aware that hypothermia, cardiac anomalies, partial pressure of carbon dioxide, pH, and haemoglobin levels can affect the accuracy of pulse oximetry.^{4}
Footnotes

Source of funding: no external funding.

For correspondence: Dr N J Shaw, Neonatal Intensive Care Unit, Liverpool Women's Hospital, Crown Street, Liverpool L8 7SS, UK. Fax +44 (0)151 702 4082.
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.