Article Text

other Versions

Cerebral desaturations in preterm infants: a crossover trial on influence of oxygen saturation target range
  1. Manuel B Schmid,
  2. Reinhard J Hopfner,
  3. Susanne Lenhof,
  4. Helmut D Hummler,
  5. Hans Fuchs
  1. Division of Neonatology and Pediatric Critical Care, Department for Pediatrics and Adolescent Medicine, University Medical Center, Ulm, Germany
  1. Correspondence to: Dr Manuel B Schmid, Division of Neonatology and Pediatric Critical Care, Department of Pediatrics and Adolescent Medicine, University Medical Center Ulm, Eythstraße 24, Ulm 89075, Germany;  manuel.schmid{at}


Objective To test the hypothesis that a higher pulsoximetric arterial oxygen saturation (SpO2) target range is associated with reduced cerebral tissue oxygen desaturations from baseline during events of hypoxaemia or bradycardia.

Design Randomised crossover trial.

Setting Single tertiary care neonatal intensive care unit.

Patients Sixteen preterm infants with severe intermittent hypoxaemia or bradycardia.

Interventions SpO2 target was set to 80–92% and 85–96% for 4 h each in random sequence. On a subsequent day, the target sequence was reversed and the study was repeated.

Main outcome measures We simultaneously recorded cerebral tissue oxygen saturation (cerebral StO2), SpO2 and heart rate. Cerebral StO2 was measured by near infrared spectroscopy. The primary outcome was the cumulative cerebral StO2 desaturation score representing the area below a cerebral StO2 baseline value before onset of each hypoxaemic or bradycardic event.

Results During low SpO2 target range the median (IQR) cumulative cerebral StO2 desaturation score was higher (27384 (15825–37396) vs 18103 (6964–32946), p=0.011) and the mean (±SD) number of events was higher (29.1 (±15.3) vs 21.1 (±11.4), p=0.001). More time was spent with SpO2 below 80% (57.2 (±24.8) min vs 34.0 (±29.6) min, p=0.006). Total time of hyperoxaemia (defined as SpO2 ≥97% and ≥99%, respectively) and total time with cerebral StO2 <60% and <55% were similar.

Conclusions A lower SpO2 target range was associated with a greater cumulative cerebral StO2 desaturation score, caused by more frequent SpO2 desaturations. However, time at very low cerebral StO2 was not affected. Episodes of hyperoxaemia were not reduced.

Statistics from

What is already known on this topic?

  • The optimal pulsoximetric arterial oxygen saturation (SpO2) target range for preterm infants is unknown.

  • A lower SpO2 target range may be associated with increased incidence of intermittent hypoxaemia.

  • There is insufficient data on the impact of SpO2 target range on cerebral oxygenation.

What this study adds

  • A lower SpO2 target range was associated with more cerebral desaturations and episodes of arterial hypoxaemia.

  • It did not reduce incidence of hyperoxaemia.

  • However, neither time at low cerebral saturation nor median cerebral oxygen saturation was affected.


Events of intermittent hypoxaemia and bradycardia affect almost all neonates of less than 30 weeks gestational age1 ,2 as a consequence of apnoea due to immature respiratory control, obstructive apnoea and loss of functional residual capacity. The peak incidence occurs at the age of 4–5 weeks.2 The rate and severity of intermittent hypoxaemia is associated with retinopathy of prematurity (ROP) as well as with neurodevelopmental impairment.2–4 Treatment options include application of continuous positive airway pressure,5 non-invasive ventilation,6 methylxanthine treatment,7 and general measures like temperature control and positioning to maintain upper airway patency.1 ,5 ,6 ,8 Some infants require endotracheal intubation and mechanical ventilation.

The optimal SpO2 target for preterm infants is still unknown and neonatologists have to balance risks of oxygen-derived toxicity associated with high pulsoximetric arterial oxygen saturation (SpO2) levels against risks of mortality and brain damage associated with low saturation levels. One mechanism by which higher SpO2 targets prevent brain damage could be that a higher SpO2 baseline may decrease the incidence of intermittent hypoxaemia and bradycardia by improving respiratory drive and thus reducing apnoeic spells.9–13 However, this finding could not be confirmed by others.14 The aim of this study was therefore to evaluate, whether a higher SpO2 target is associated with decreased incidence and severity of cerebral desaturation during events of hypoxaemia and/or bradycardia. We used cerebral tissue oxygen saturation (cerebral StO2) as our primary outcome criterion because it reflects the effects of poor oxygenation as well as the effects of poor perfusion, for example, during low cardiac output states related to severe bradycardia. Adequate brain oxygenation could be more important to prevent brain damage than SpO2.



We included preterm neonates less than 34 weeks gestation at the time of birth and actual age less than 40 weeks gestation with severe intermittent hypoxaemia or bradycardia requiring supplemental oxygen and treated in our neonatal intensive care unit between September 2010 and April 2011. Severe intermittent hypoxaemia or bradycardia was defined after discussion among the investigators as ≥8 events of SpO2 <75% or bradycardia <80 bpm within the last 8 h. We excluded infants with congenital malformations, neuromuscular diseases or diaphragmatic paralysis. Imaging studies with head ultrasound including Doppler examination were performed on all participants according to hospital standards.

Treatment of apnoea of prematurity and immediate reactions to events of apnoea, hypoxaemia or bradycardia were guided by a local hospital guideline.


Patients were randomised to Protocol A or Protocol B by closed envelopes (figure 1). In protocol A the SpO2 target was set for 4 h to 80–92%, which has been the standard target range in our neonatal intensive care unit in preterm infants for many years, followed by 85–96% for another 4 h (Study Day 1). On a subsequent day the recording was repeated with reversed target sequence (Study Day 2). Protocol B was identical but the SpO2 target sequence was reversed as compared with protocol A.

Figure 1

Patient flow diagram.

Outcome parameters

Heart rate as determined by ECG and preductal SpO2 were extracted from the bedside monitor (Dash 5000, GE Medical Systems, Munich, Germany; Masimo SET Firmware version for pulse oximetry V4.3.1.0). SpO2 averaging time was 2 s, sensitivity was set to ‘normal’. Cerebral StO2 was measured by near infrared spectroscopy (NIRS) using absolute oximetry; a laser-based sensor using four different wavelengths for cerebral StO2 monitoring (FORE-SIGHT, Casmed, Branford,  Connecticut, USA) was attached at both sides of the infant's head. Sampling interval for all parameters was 2 s. These data were collected simultaneously on a laptop computer using dedicated software (LabVIEW, National Instruments Corporation, Austin, Texas, USA). Averaged cerebral StO2 values of left and right sensors were calculated. Fractional tissue oxygen extraction (FTOE) was calculated as 100*(SpO2−cerebral StO2)/SpO2. Haemoglobin difference (HbD) was calculated as oxygenated haemoglobin−deoxygenated haemoglobin and is regarded as an index for perfusion. These postprocessed calculations and graphical visualisation of all parameters were done with Microsoft Excel 2010 (Microsoft, Redmond, Washington, USA).

The primary outcome measure was the total cerebral StO2 desaturation, defined as a cumulative cerebral tissue desaturation score of all events resulting from hypoxaemic or bradycardic events. An event was defined as SpO2 <75% or ECG heart rate <80 bpm. A baseline value representing the last stable cerebral StO2 value prior to onset of the event was determined individually for every event. The difference between actual cerebral StO2 and baseline was calculated every 2 s until StO2 returned to the threshold value (area below the baseline threshold, figure 2). Thereby, every StO2 desaturation during an event was counted, regardless of its extent. The sum of these differences is a score reflecting the length and depth of each event, and the cumulative score is the sum of scores for all events. StO2 changes in the absence of SpO2 desaturations or bradycardias were not considered.

Figure 2

Example of an apnoeic spell with bradycardia and subsequent hypoxaemia. Cerebral StO2 baseline value in this example is 78%, area under the threshold is calculated until return to the baseline (primary outcome criterion).

Secondary outcome measures were number of events, score per event, total time with SpO2 <80% and <75% as well as ≥97% and ≥99%, time with heart rate <100 bpm and <80 bpm, time with cerebral StO2 <60% and <55%, time within the 10th to 90th and >90th and >97th percentile for cerebral StO2. Cerebral StO2 percentiles were determined using all cerebral StO2 values where SpO2 was between 85% and 92%, the heart rate was between the 3rd (131 bpm) and the 97th (183 bpm) percentile, and infants did not receive catecholamines or sedatives, and did not have intracranial haemorrhage.


A power analysis for sample size could not be calculated, as the variability of the primary outcome criterion was not known. For practical reasons, we selected a sample size of 16 patients. Statistical analysis to compare paired data was done with SigmaPlot (Systat Software, San Jose, California, USA) using paired t-test for parametric data and Wilcoxon Signed Rank Test for non-parametric data. A p-value <0.05 was considered statistically significant. Data from both study days were summed.

The study was approved by the local ethics committee (Ethikkommission Ulm University 184/10). All parents have given informed consent.


Characteristics of study infants are given in table 1. No sedatives were given except phenobarbital at a maximum dose of 5 mg/kg/d. All infants were on caffeine.

Table 1

Patient characteristics and baseline parameters on Study Day 1 and Study Day 2 including last haematocrit, actual dose of caffeine, mode of respiratory support, supplemental oxygen and number of patients which are already set on elevated SpO2 target range for clinical reasons

Signal availability (median, IQR) for heart rate (99.9%, 99.9–100.0), SpO2 (99.7%, 99.6–99.9) and cerebral StO2 (98.1%, 94.7%–99.8%) was high and did not differ between target range phases. All patients completed the study.

Main results are given in table 2. No adverse effects were observed.

Table 2

Main results during low (80–92%) and high (85–96%) target range phases. Parametric data are tested with paired t-test, mean and SD are given; nonparametric data are tested with Wilcoxon Signed Rank Test

Median SpO2 and quartiles were significantly higher in the high SpO2 target range phase. During the low SpO2 target range phase, infants spent less time within the intended target range, more time with SpO2 <80% and <75%, whereas there was no significant difference for time with SpO2 ≥97% and ≥99% (figure 3C).

Figure 3

Boxplot diagrams showing cumulative cerebral desaturation score, number of events and cerebral desaturation score per event (A), medians of SpO2, cerebral StO2, FTOE, HbD and total haemoglobin of the whole recording time (B) and time spent below and above arterial (C) and cerebral saturation thresholds (D). Grey bars represent low target range, white bars high target range. Boundaries of the boxes indicate 25th and 75th percentiles, whiskers indicate 10th and 90th percentiles, and dots indicate outliers. A line in the box indicates the median, a dashed line the mean value. An asterisk indicates a p value below 0.05 for difference between medians. Cerebral StO2, cerebral tissue oxygen saturation; FTOE, fractional tissue oxygen extraction; HbD, Haemoglobin difference; SpO2, pulsoximetric arterial oxygen saturation.

The median cumulative cerebral StO2 desaturation score and mean number of events were significantly higher during the low SpO2 target range phase with no significant difference in the score per event (table 2, figure 3A). During the low target range phase infants spent slightly more time with cerebral StO2 <60% and <55% as well as >90th and >97th, respectively, but these differences were not significant (figure 3D). The 3rd, 10th, 90th and 97th percentiles of cerebral StO2 were 53.0, 59.0, 81.0 and 84.5, respectively. Time of cerebral StO2 within the 10th and 90th percentiles did not differ between groups, nor did median cerebral StO2, FTOE, total haemoglobin concentration or HbD (figure 3B). There were undulating changes of cerebral StO2 but significant StO2 desaturations in the absence of SpO2 events or heart rate events were not observed. Frequency histograms of SpO2, cerebral StO2, cerebral FTOE and cerebral HbD are shown in figure 4. The histogram of SpO2 and consequently of FTOE shows a dip between SpO2 of 87% and 90%, which appears more pronounced during the low SpO2 target range phase.

Figure 4

Frequency histograms of SpO2, cerebral StO2, HbD, and FTOE. Time spent at discrete values during high (dashed line) and low (continuous line) target range phase. Cerebral StO2, cerebral tissue oxygen saturation; FTOE, fractional tissue oxygen extraction; HbD, Haemoglobin difference; SpO2, pulsoximetric arterial oxygen saturation.


Cerebral oxygenation and cerebral blood volume as measured by NIRS have been shown to be influenced by apnoea and hypoxaemia before.15–24 However, all of these studies deal with immediate influences of apnoeic or desaturation episodes and this is, to our knowledge, the first study to evaluate effects of different SpO2 targets on cerebral oxygenation and it is the first study using absolute cerebral tissue oximetry evaluating hypoxaemic events.

We examined 16 preterm infants at a median age of 5 weeks, where incidence of intermittent hypoxaemic episodes is highest.11 Our definition for an event of intermittent hypoxaemia or bradycardia differs from the recommendation for definition of apnoeic spells made by the apnoea-of-prematurity group:25 For definition of an intermittent hypoxaemia event we have chosen a SpO2 limit of 75% instead of 80% because the lower limit of the lower target range was 80%. Using this limit as a threshold could have resulted in a high incidence of events of little clinical significance in the lower SpO2 target range phase. For bradycardic events, we adapted the limit of 80 bpm as suggested by Finer et al.25 However, the heart rate baseline of every single infant was higher and we never observed a stable heart rate with values below 120 bpm, suggesting that a threshold of 120 bpm or relative to an individual baseline may be more appropriate to indicate significant bradycardia.26

Little is known about the effects of short but frequent events of hypoxaemia and hypoperfusion on the brain of preterm infants. In asphyxia, duration of hypoxaemia and ischaemia is related to the magnitude of brain damage. Large comparative observative studies show no difference in cognitive development between firstborn children, who experience significantly more episodes of bradycardia and desaturation during spontaneous birth compared with their younger siblings. This supports the hypothesis, that there may be a ‘safe’ duration limit below which hypoxaemia and hypoperfusion are harmless, but we do not know this threshold. Thus we used NIRS to evaluate the influence of intermittent hypoxaemia and bradycardia upon cerebral oxygenation. As there are no established thresholds for safe or ‘normal’ cerebral tissue oxygenation, we decided to use deviation from a stable baseline value before onset of an event as primary outcome. Our cerebral tissue desaturation score reflects the area under the threshold for every event, taking into account the duration and severity (depth) of the cerebral desaturation. We found a significantly higher score during the low target range phase. The difference was mainly due to a higher number of events in the low target range phase whereas the desaturation score per event was only slightly higher. This supports the hypothesis of improved respiratory stability at higher SpO2 levels. As the definition of an event in our study was solely based on SpO2 desaturation and/or bradycardia, our primary outcome is necessarily largely influenced by the frequency of SpO2 desaturations and bradycardia.

Despite the higher incidence of cerebral desaturations and a longer duration at low SpO2 values during the low SpO2 target phase we did not find significant differences in the time spent at low cerebral StO2 values. This could be explained by compensatory mechanisms like increased cerebral blood flow during hypoxaemia.

Despite significant differences in SpO2, the mean and median cerebral StO2 during total recording time did not differ between phases. This could be a mathematical artefact: The frequency distribution histogram suggests that the increased incidence of lower (60–70%) cerebral StO2 values is balanced by an increased incidence of high values (82–92%) resulting in a similar average. Another explanation would be a compensatory increase of cerebral perfusion or a decrease in oxygen consumption during the low target range phase. However, the perfusion markers total haemoglobin concentration and HbD were not increased during low target range. The lower FTOE in the low target range suggests decreased oxygen consumption associated with lower arterial oxygen content. However, as these are only trends, this finding should be interpreted cautiously. FTOE always requires cautious interpretation as it is calculated from two values with limited reliability: There is a considerable difference between SpO2 and arterial oxygen saturation.27 Saturation is a poor predictor of arterial oxygen tension, especially for infants receiving supplemental oxygen and for high saturation values.1 ,2 ,28 These limitations are likely to apply for StO2 measurements too.

There are no established reference values for cerebral StO2 as determined by NIRS. We calculated percentiles from our small cohort of infants at the age of 4–5 weeks, determined in the absence of hypoxaemia or bradycardia during a SpO2 level of 85–92%. This corresponds to the intersection of the two target ranges used in our study and is close to the low SpO2 target range of SUPPORT29 and BOOST II30 trial. These percentiles are probably different at different SpO2 targets. Furthermore, they could differ if different StO2 devices are used and thus should not be generalised.

One rationale to use a low SpO2 target is to avoid hyperoxaemia and its clinical sequelae such as ROP. However, time at SpO2 ≥97% or ≥99% was not different between the two target ranges. Times at cerebral StO2 >90th and >97th percentiles were longer during the low target range, but the differences were not statistically significant. It is likely that overregulation by caregivers by increasing oxygen in response to more severe and frequent hypoxaemia was more common in the low target range phase. This could explain at least in part the association of frequent hypoxemic events and ROP observed by others.31 The increased incidence of high cerebral StO2 is in concordance with the findings of Baerts et al,23 who found a reduced FTOE and an increased cerebral StO2 in association with overshooting oxygen regulation for desaturations.

Time spent with bradycardia below 80 bpm or 100 bpm, respectively, did not differ between the target ranges, indicating that the additional hypoxaemias during low target range may not have been severe enough to cause bradycardia.

Our study has certain limitations. First, as this was an unblinded study, caregivers might have been influenced. However, we believe that there is a low risk for bias, as we did not encounter any attitude changes during the different target ranges among caregivers. Furthermore, SpO2 target ranges were fairly adhered to.

Second, we recognised a gap in the frequency distribution histogram of SpO2. A similar finding was observed in the BOOST II trial30 and is attributed to an error of the calibration algorithm of the Masimo SET pulse oximeter prior to firmware V4.8, causing an artefactual elevation of SpO2 readings between 87% and 90%.32 The Masimo pulse oximetry module of the GE Dash monitors used in our unit uses a firmware version that is affected by the artefact. This gap is affecting oxygen targeting particularly in our lower target range: if SpO2 values of above 90% are shown with an increased frequency caregivers will reduce supplemental oxygen, possibly resulting in more desaturations and more caregiver interventions. Use of a SpO2 target of 80–92% in combination with the algorithm affected by the calibration artefact may therefore predispose to more desaturations.

Third, inclusion criteria were liberal resulting in a wide range of postnatal age, postmenstrual age, mode of respiratory support and haematocrit. This may have diluted differences that exist in certain subgroups. Unfortunately our trial is too small to perform subgroup analyses.

Fourth, we did not attempt to measure respiratory efforts or nasal gas flow due to technical limitations. Therefore, the impact of SpO2 target on respiratory centre output could not be determined. It is possible that inclusion of hypopnoea or apnoea without hypoxaemia or bradycardia could have altered our findings.

Fifth, SpO2 targets used in our study are below SpO2 targets used in most hospitals. We can only speculate about differences between two higher SpO2 target ranges.

The best SpO2 target range for preterm neonates is a matter of ongoing debate.2–4 33–35 It may differ between subjects, depending on gestational age, postnatal age and medical condition.2 ,5 ,35 ,36 Recent studies emphasise the impact of SpO2 target range on mortality6 ,29 ,30 and morbidity of preterm infants. In particular, the incidence of ROP,7 ,29 ,30 ,34 ,37–39 bronchopulmonary dysplasia,1 ,8 ,29 necrotising enterocolitis,29 intermittent hypoxaemia or apnoea of prematurity9–13 ,29 ,40 and cerebral palsy2 could be affected by SpO2 targets. Our study supports the hypothesis of an association between a low SpO2 target, increased incidence of hypoxemic events and more cerebral desaturations.


In infants with intermittent SpO2 desaturations and bradycardias, cerebral desaturations were more frequent in the lower SpO2 target range phase (80–92%), as compared with the higher target range phase (85–96%). This was caused by a higher number of SpO2 desaturations. The lower target range did not prevent episodes of arterial or cerebral tissue hyperoxaemia. Time at very low cerebral StO2 values, however, was not significantly different. Further studies are needed to evaluate the impact of cerebral desaturations on clinical outcomes and to help determine a normal cerebral StO2 range in this population of preterm infants.


We thank Bob Kopotic and Roxana Mois for their technical support.

View Abstract


  • Contributor MBS was the principal investigator of the study and the author of the manuscript. RJH and SL collected and evaluated the data and critically reviewed the manuscript adding valuable information. HDH supervised and was involved in the development of the protocol, data collection and evaluation. He had principal influence on the manuscript. HF also was involved in development of the protocol, data acquisition and he coauthored the manuscript.

  • Competing interests None.

  • Ethics approval Ethikkommission Uni Ulm.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Data on the influence of different types of events on cerebral oxygenation is available and is still being evaluated. It will be submitted separately by the same group.

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.