Article Text

Download PDFPDF

Burden of hypoxia and intraventricular haemorrhage in extremely preterm infants
  1. Isabel Hui Xuan Ng1,
  2. Cristine Sortica da Costa1,
  3. Fred A Zeiler2,3,
  4. Flora Y Wong4,
  5. Peter Smielewski5,
  6. Marek Czosnyka5,
  7. Topun Austin1
  1. 1 Neonatal Unit, Rosie Hospital, Cambridge University Hospitals NHS Foundation Trust, Cambridge, UK
  2. 2 Department of Medicine, Addenbrooke’s Hospital Department of Anaesthesia, Cambridge, UK
  3. 3 Section of Neurosurgery, Department of Surgery, Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, Manitoba, Canada
  4. 4 The Ritchie Centre, Monash University, Monash Institute of Medical Research, Clayton, Victoria, Australia
  5. 5 Department of Clinical Neurosciences, University of Cambridge, Cambridge, UK
  1. Correspondence to Dr Topun Austin, Neonatal Intensive Care Unit, Rosie Hospital, Cambridge University NHS Foundation Trust, Cambridge Biomedical Campus, CB2 0QQ, UK; topunaustin{at}


Objective Thresholds of cerebral hypoxia through monitoring of near-infrared spectroscopy tissue oxygenation index (TOI) were used to investigate the relationship between intraventricular haemorrhage (IVH) and indices of hypoxia.

Design Prospective observational study.

Setting A single-centre neonatal intensive care unit.

Patients Infants <28 weeks’ gestation with an umbilical artery catheter.

Methods Thresholds of hypoxia were determined from mean values of TOI using sequential Χ2 tests and used alongside thresholds from existing literature to calculate percentage of time in hypoxia and burden of hypoxia below each threshold. These indices were then compared between IVH groups.

Results 44 infants were studied for a median of 18.5 (range 6–21) hours in the first 24 hours of life. Sequential Χ2 analysis yielded a TOI threshold of 71% to differentiate between IVH (16 infants) and no IVH (28 infants). Percentage of time in hypoxia was significantly higher in infants with IVH than those without, using thresholds of 60%–67%. Burden of hypoxia was significantly higher in infants with IVH than without, using thresholds of 62%–80%. With the threshold of 71%, percentage of time in hypoxia was lower by 12.2% with a 95% CI of (−25.7 to 1.2) (p=0.073), and the burden of hypoxia was lower by 29.2% hour (%h) (95% CI −55.2 to −3.1)%h (p=0.012) in infants without IVH than those with IVH.

Conclusions Using defined TOI thresholds, infants with IVH spent higher percentage of time in hypoxia with higher burden of cerebral hypoxia than those without, in the first 24 hours of life.

  • neonatology
  • neurodevelopment
  • neurodisability
  • monitoring
  • imaging

Statistics from

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.

What is already known on this topic?

  • Cerebral hypoxia below a threshold has been associated with adverse outcomes in animal and human infant studies.

  • Near-infrared spectroscopy is a non-invasive method of continuous measurements of cerebral oxygenation.

  • Thresholds are important as clinical targets.

  • Burden of cerebral hypoxia in preterm infants may be lowered through cerebral oxygenation monitoring using near-infrared spectroscopy and treatment guidelines targeting cerebral oxygen delivery.

What this study adds?

  • Thresholds of hypoxia may be defined in extremely preterm infants in the first 24 hours, where a significant association with intraventricular haemorrhage (IVH) is observed.

  • Infants with IVH spend significantly higher percentage of time in cerebral hypoxia than those without IVH.

  • Thresholds identified in this study are higher than those investigated in existing literature, which may be accounted for by between-oximeter variation.


Preterm birth remains a leading cause of neonatal mortality and morbidity. Prematurity is associated with lifelong neurodevelopmental disability.1 2 Extremely preterm infants (<28 weeks) are at risk of intraventricular haemorrhage (IVH) in the first days of life, which has been linked to cerebral palsy and intellectual impairment.3 Decreased cerebral blood flow and oxygenation is associated with brain injury, death and/or IVH.4–6 Therefore, continuous monitoring of cerebral blood flow and oxygenation is of clinical relevance in identifying and managing brain injury.

Near-infrared spectroscopy (NIRS) provides a safe, non-invasive measurement of cerebral oxygenation, expressed as percentage tissue oxygen saturation, or tissue oxygenation index (TOI). Identifying thresholds of cerebral oxygenation are important in the clinical setting as targets that could be integrated into treatment guidelines. Thresholds associated with adverse outcomes have been observed in animal and human models.7–10 Large population studies of NIRS measurements in preterm infants also provide information about population values of cerebral oxygenation.11 The multicentre randomised controlled trial, Safeguarding the Brains of our Smallest Children (SafeBoosC), used a threshold of cerebral hypoxia of 55%, based on data measured using the INVOS 5100C oximeter and the Adult SomaSensor (Invos adult sensor) (Somanetics-Covidien, Dublin, Ireland).12 They described the calculation of burden of cerebral hypoxia, which takes into account both magnitude and duration of deviation below this threshold.12 The phase II trial demonstrated that burden of cerebral hypoxia in preterm infants may be lowered through cerebral oxygenation monitoring and treatment guidelines targeting cerebral oxygen delivery.13 To our knowledge, there have been no previous data published evaluating the burden of cerebral hypoxia using a range of thresholds and its relationship to adverse outcomes in preterm neonates.

In this study, we aimed to assess the relationship between IVH and different thresholds of cerebral hypoxia in the first 24 hours of life, using the NIRO-200NX oximeter with a small reusable sensor (Hamamatsu Photonics, Hamamatsu City, Japan). We investigated the association between IVH and mean TOI values, percentage of time spent in cerebral hypoxia and burden of cerebral hypoxia.


Patient population

This was a single-centre prospective observational study of preterm infants born at <28 weeks’ gestational age who had indwelling arterial catheters inserted for clinical reasons. Infants born with major malformations were excluded. The recruitment protocol was the same applied for a previous study on cerebrovascular reactivity published by our group.14 The infants included are the same as a previous study on optimal blood pressure published by our group.15 The median (range) age of the 44 infants at the start of the study was 5.5 (3.1–12.6) hours and the median (range) duration of study was 18.5 (6.5–20.9) hours.

Data acquisition and processing

The infants were studied using a NIRO-200NX NIRS monitor. The NIRO-200NX measures changes in the concentration of oxyhaemoglobin (HbO2)and haemoglobin and provides measurements of TOI. The NIRO-200NX uses three light-emitting diodes with wavelengths of 735, 810 and 850 nm and two detecting photodiodes to measure light attenuation at different distances from the source. A sensor holder was used to fix the source at 3 cm away from the detector. A light-proof cover was used. The sensor was changed to the opposite temporoparietal side of the infant’s head every 6–8 hours to avoid skin marks. Continuous TOI data were collected and retrospectively analysed using ICM+ software (Cambridge Enterprise, Cambridge, UK), including manual artefact removal.16 Only data collected within the first 24 hours of age were included in the analysis. Clinical data were collected from documented patient notes. IVH was defined according to Papile et al 17 and the greatest grade of haemorrhage observed during admission was used for analysis.

Thresholds of cerebral hypoxia

Thresholds of cerebral hypoxia were determined from mean values of TOI in the first 24 hours of life, using sequential Χ2 testing, a methodology applied to studies in adults with traumatic brain injury to identify thresholds of indices of cerebral autoregulation.18–20 Sequential Χ2 tests were performed to test thresholds of TOI for their association with the outcome of IVH. Sequential 2×2 contingency tables were created where patients were grouped based on presence versus absence of IVH and mean TOI values above versus below sequential thresholds (moving in 1% increments of TOI). The Χ2 test statistic value was then plotted against the TOI thresholds tested. The value of TOI with the largest statistically significant (p<0.05) Χ2 value was deemed the threshold.

From the literature, Alderliesten et al 11 demonstrated 5th and 95th percentiles of regional cerebral oxygen saturations of 60%–65% and 80%–87%, respectively in 660 infants of gestational ages 24–29 weeks over the first 24 hours of age.11 The SafeBoosC trial used 55% as their threshold of hypoxia based on unpublished data of preterm infants, where 55% was 2 SD away from the mean NIRS readings from 390 infants.12 The thresholds used both in the SafeBoosC trial (55%) and reference values from the study by Alderliesten et al (60%–87%) were investigated in conjunction with the threshold from our Χ2 analysis to calculate burden of hypoxia and percentage of time spent below each threshold.

Indices of hypoxia

Percentage of time in hypoxia was calculated as the duration of TOI measurements (in 10 s averages) below the threshold, as a percentage of overall study duration.

For burden of hypoxia (%h), magnitude of deviation was calculated by considering each 10 s window and if TOI measurement (%) was below the threshold, the difference between the TOI measurement and the threshold was calculated. Burden was then calculated using the mean magnitude of deviation (%), multiplied by the duration of TOI measurements below the threshold (in hours).12

Python programming software (Python, V.3.6.1) was used for the calculations of the indices of hypoxia.

Statistical analysis

Data were assessed for normality using the Shapiro-Wilk test. Depending on the data distribution, parametric (independent samples t-test) or non-parametric (Mann-Whitney U test) tests were applied to compare between infants with and without IVH, for the percentage of time spent below the threshold of cerebral hypoxia and burden of cerebral hypoxia. Means and SD were used to describe clinical characteristics of outcome groups. Chi-squared test was used to compare differences in binary variables between outcome groups. A statistical test was considered significant if the p value was <0.05 (two-tailed). Statistical analysis was performed using SPSS software package V.23 (SPSS, Chicago, Illinois, USA).


Sample population

Table 1 shows the characteristics of infants in each outcome group. Sixteen infants developed IVH (grades 1–4), among whom one developed a grade 1 haemorrhage; six infants, grade 2; five infants, grade 3 and four infants, grade 4. In four infants, the presence of an IVH was observed in the cranial ultrasound performed before 24 hours of life (25%). In the remaining infants who developed IVH, bleeding was observed on the cranial ultrasound performed at 24 hours (31%) or later (44%). None of the infants included died during NIRS data collection, but six died before 40 weeks of corrected gestational age. Mean (SD) TOI (%) was significantly lower in the IVH group compared with no IVH: 72.6 (4.07) vs 70.3 (5.82) (p=0.04).

Table 1

Baseline characteristics of the study population in outcome groups

Determination of thresholds

Chi-squared analysis of mean TOI and IVH identified significant Χ2 values (p<0.05) with thresholds of 69%–71%, as shown in table 2 and figure 1, with a threshold of 71% having the largest Χ2 value. Together with the threshold used in the SafeBoosC trial and reference values found by Alderliesten et al,11 we investigated thresholds of 55%, 60%, 62%, 65%, 67%, 70%, 71%, 72%, 75%, 77%, 80%, 82%, 85% and 87%.11 12

Table 2

Number of infants with or without IVH with a mean TOI (%) below the given threshold, percentage of total shown in brackets

Figure 1

Chi-squared analysis of thresholds of mean TOI (%) against outcome of intraventricular haemorrhage. The Χ2 value for each threshold has been plotted against the TOI (%). The threshold with the largest significant Χ2 value has its p value marked on the graph—a threshold of 71%. IVH, intraventricular haemorrhage; TOI, tissue oxygenation index.

Analysis between outcome groups

Percentage of time of cerebral hypoxia

Table 3 shows the difference between mean percentage of time of cerebral hypoxia between IVH groups. Infants who developed IVH had higher percentage of time with thresholds of 60%, 62%, 65% and 67% (p<0.05).

Table 3

Comparison of percentage of time spent below threshold of cerebral hypoxia (%) with 95% CIs between outcome groups

Burden of cerebral hypoxia

Table 4 shows the difference between mean burden of cerebral hypoxia between IVH groups. Infants who developed IVH had a higher mean burden with thresholds of 62%, 65%, 67%, 70%, 71%, 75%, 77% and 80% (p<0.05).

Table 4

Comparison of mean burden (%h) of cerebral hypoxia with 95% CIs (%) between outcome groups


This study assessed the relationship between deviation of TOI below thresholds and the outcome of IVH. IVH has been demonstrated to mostly occur in the first 72 hours of life. By using data from the first 24 hours of life in this analysis, we aimed to identify cerebral hypoxia within the high-risk period for developing IVH. In our patient population, infants with IVH were noted to have significantly lower mean TOI than infants without IVH, although the absolute difference was small (73% and 70%).

Infants with IVH were found to spend a significantly higher percentage of time in hypoxia than infants without IVH when thresholds of 60%–67% were used. Furthermore, the percentage of time in hypoxia had a narrower and lower range of significant thresholds than those for the burden of hypoxia. This highlights the importance of time spent with cerebral oxygenation below these lower thresholds, regardless of the magnitude of hypoxia.

Previous studies have investigated lower thresholds of cerebral hypoxia than observed in our study. Cerbo et al investigated thresholds of hypoxia of 40% and 55% based on animal studies. They found that a threshold of 40%, but not 55%, led to a significant difference in the ‘count of cerebral oxygenation measurements’ below the threshold between groups of infants based on outcomes of mortality and severe IVH.7 However, they did not investigate thresholds >55%.

Published data from the SafeBoosC trial, which uses a threshold of 55%, found that it was possible to significantly reduce burden of cerebral hypoxia through management of cerebral oxygen delivery using an evidence-based treatment guideline.12 The trial showed significantly better cranial ultrasound severity scores in the treatment group than the control group in early scans and at term equivalent age. However, there was no significant difference in overall cranial ultrasound scores, MRI findings, early electroencephalography, blood biomarkers of brain injury or neurodevelopmental outcomes between their two trial groups.21–23 Our study highlights the possibility that the threshold of 55% is not applicable to all study populations. In our study population, no difference in the incidence of IVH was associated with a threshold of 55% and the infants had very few TOI measurements <55%.

NIRS system-related variation in cerebral oxygenation readings may account for different thresholds observed. There are several different NIRS systems in use, with both adult and paediatric sensors. Alderliesten et al used an INVOS 4100 or 5100(c) monitor with an adult sensor to create their reference curves.11 In their study, they described a conversion model to convert neonatal sensor readings to an adult sensor equivalent. They reported a difference between the neonatal and adult sensors of up to 15% (mean 10%), with the neonatal sensor reading higher.

A study comparing various NIRS oximeters on a liquid phantom by Kleiser et al modelled transformation of cerebral oxygenation values between various oximeters. Using the hypoxic threshold of 55% used in the SafeBoosC trial and based on the measurements using the adult INVOS sensor, the model by Kleiser et al suggests an equivalent hypoxic threshold of 61.0%, with a range of uncertainty of 3.5% measured by the NIRO-200NX oximeter using a small sensor.24 This could explain the higher thresholds identified in our study. In our study using the NIRO-200NX oximeter with a small sensor, thresholds of 60%–67% were found to be significant when investigating percentage of time spent in hypoxia between infants with and without IVH, which is similar to the threshold of 61% for the NIRO-200NX oximeter suggested by Kleiser et al.24

The difference between the thresholds of cerebral hypoxia identified in our study and those used in the existing literature highlights the difficulty with defining universal thresholds as these may be population-dependent and equipment-dependent. The SafeBoosC threshold of 55%, using the INVOS adult sensor, to prompt correction of cerebral hypoxia, may be too low for our study population. The suboptimal threshold could have led to significant cerebral hypoxia in the infants even before treatment is initiated, resulting in reduced benefit from the treatment. Further investigation into thresholds and adverse outcomes in a larger sample is warranted to understand the role of NIRS as a clinical tool. Thresholds should also be adjusted for specific NIRS systems given between-system variation.


Despite using artefact-removed continuous NIRS data, the small sample size of this study means that we should approach the thresholds found in this study with caution. It is possible that the presence of IVH may lead to TOI readings that do not reflect global cerebral oxygenation and blood flow, or indeed itself lead to impaired cerebral autoregulation and cerebral hypoxia.25 We minimised this potential confounding variable by using data from the first 24 hours, when most infants had yet to develop IVH.

There was no significant difference in mean blood pressure, peripheral oxygen saturation or inotrope use between infants with or without IVH. However, there are additional contributing factors to cerebral hypoxia that were not investigated, such as PaO2 of carbon dioxide, and it is therefore difficult to comment on cause of hypoxia.

Our data suggest that periods of cerebral hypoxia with even mild deviation below 62% may be associated with IVH, as the percentage of time spent in hypoxia was significantly different between infants with and without IVH, regardless of magnitude. However, our analysis was not designed to differentiate between the relative contribution of duration and magnitude to the significant difference observed when cerebral oxygenation is below 62%. In other words, it is not certain how the magnitude of cerebral hypoxia below 62% adds to the adverse effects from the duration of cerebral hypoxia.


Continuous non-invasive monitoring using NIRS can provide cot-side, real-time information on cerebral hypoxia and hypoperfusion. Our study demonstrates that objective thresholds of early cerebral hypoxia may be defined, below which an association with IVH is observed. The thresholds of hypoxia, where percentage time spent in hypoxia was found to be associated with IVH, are similar to those described in existing literature when between-oximeter variation is taken into consideration. Our study highlights the potential impact of prolonged periods of cerebral hypoxia and the need for further investigation into thresholds of cerebral hypoxia to guide the use of NIRS in clinical management. With defined thresholds of cerebral hypoxia, bedside monitoring and interventions to reduce cerebral hypoxia may reduce risk of IVH in preterm infants.



  • Contributors Substantial contributions to the conception or design of the work: CSdaC, PS, MC and TA. Acquisition, analysis or interpretation of data for the work: IHXN, CSdaC and FAZ. Drafting the article: IHXN. Revising the article critically for important intellectual content and final approval for publication: all authors.

  • Funding CSdaC was supported by SPARKS charity (11CUH02), the Cambridge Trust and the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) (PhD scholarship to Dr Sortica da Costa/9418–11–3). FAZ is supported by the University of Manitoba Thorlakson Chair in Surgical Research Establishment Grant, University of Manitoba VPRI Research Investment Fund (RIF), Winnipeg Health Sciences Centre (HSC) Foundation and the University of Manitoba Rudy Falk Clinician-Scientist Professorship.

  • Competing interests The ICM+ software (ICM+; used for data monitoring and analysis is licensed by Cambridge Enterprise Limited (University of Cambridge). PS and MC have an interest in a fraction of the licensing fee.

  • Ethics approval The study was authorised by The Research and Development Department of Cambridge University Hospitals NHS Foundation Trust and approved by The East of England Research Ethics Committee (12/EE/0524).

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

  • Patient consent for publication Not required.