Article Text

Download PDFPDF

A cohort study reporting normal oximetry values in healthy infants under 4 months of age using Masimo technology
  1. Hazel J Evans1,
  2. Anne S Karunatilleke2,
  3. Sarah Grantham-Hill1,2,
  4. Johanna C Gavlak1
  1. 1 Department of Respiratory Paediatrics, Southampton Children’s Hospital, Southampton University Hospital NHS Foundation Trust, Southampton, Hampshire, UK
  2. 2 Faculty of Medicine, University of Southampton, Southampton, UK
  1. Correspondence to Dr Hazel J Evans, Department of Respiratory Paediatrics, Southampton Children’s Hospital, Southampton University NHS Trust, Southampton SO16 6YD, UK; hazel.evans{at}


Objective To determine sleeping saturation indices in healthy infants using a modern pulse oximeter with motion artefact extraction technology.

Design Prospective cohort.

Setting Home.

Subjects Healthy term infants.

Intervention Nocturnal pulse oximetry at home at 1 month of age (Recording 1) and repeated at age 3–4 months (Recording 2). Parents documented sleep times. Visi-Download software (Stowood Scientific) analysed data with artefact and wake periods removed.

Main outcome measures Saturations (SAT50), desaturation index >4% (DI4) and >3% (DI3) from baseline/hour, delta index 12 s (DI12s), minimum saturations (SATmin), percentage time with saturations below 90% and 92%.

Results Forty-five babies were studied at 1 month and 38 babies at 3–4 months. Mean (CI) SAT50, DI4, DI3, DI12s and SATmin (CI) were 97.05 (96.59 to 97.52), 16.16 (13.72 to 18.59), 25.41 (22.00 to 28.82), 0.96 (0.88 to 1.04) and 80.4% (78.8% to 82.0%) at 1 month, respectively, and 97.65 (97.19 to 98.12), 8.12 (6.46 to 9.77), 13.92 (11.38 to 16.47), 0.72 (0.65 to 0.78) and 84.7% (83.3% to 86.1%) at 3–4 months. Median (CI) percentage times with saturations below 90% and 92% were 0.39 (0.26 to 0.55) and 0.82 (0.60 to 1.23), respectively, at 1 month and 0.11 (0.06 to 0.20) and 0.25 (0.17 to 0.44) at 3–4 months. For paired samples (n=32) DI4 (P=0.006), DI3 (P=0.03), DI12s (P=0.001), percentage time with saturations below 90% (P=0.001) and 92% (P=0.000) all fell significantly and SATmin (P=0.004) rose between the two recordings.

Conclusion Desaturation indices are substantially higher in young infants than older children where a DI4 over 4 is considered abnormal. These decrease by 3–4 months of age but still remain elevated compared with older children.

  • respiratory

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?

  • Normative oximetry data is available for groups of children over the age of 1 year.

  • Infants are much more prone to short central apnoeas than older children.

  • Central apnoeas are known to be associated with falls in oxygen saturation.

What this study adds?

  • This study provides data on commonly reported oximetry variables used in clinical practice for very young infants using motion resistant oximeters with short averaging times.

  • This study demonstrates a shift in oximetry variables towards those quoted for older children in the first 4 months of life.

  • Four hours of data collection appears appropriate for very young infants. Defining a minimum sleep recording duration for older children requires further study.


Sleep disordered breathing (SDB) is an umbrella term to describe a group of conditions resulting in disturbed ventilation during sleep. These forms of SDB vary in mechanism but can broadly be classified into two groups; obstructive sleep apnoea (OSA) and central sleep apnoea (CSA). With the exception of congenital central hypoventilation syndrome1 2 disturbance of ventilation is uniquely associated with rapid eye movement (REM) sleep when muscle atonia predisposes to obstructive events and reduced respiratory drive contributes to central events.3 4 Unlike older children, newborn infants sleep for up to 18 hours a day and spend as much as 50% of this time in REM sleep.5 The proportion gradually reduces to 30% by the age of 1 year and 20% during adolescence.6

Instability of the breathing pattern is an inherent characteristic of normal healthy infants during sleep.7 Their unique physiology in relation to ventilatory control and susceptibility to periodic breathing make them predisposed to central apnoeas where there is a brief arrest in ventilation resulting from an interruption or decrease in respiratory drive within the central nervous system. These episodes are thought to be due to immaturity of their respiratory control mechanisms and occur predominantly during episodes of REM sleep.8

Short pauses in breathing result in transient falls in oxygen saturation.9 These can be detected by modern pulse oximeters which have short averaging times.10 They are used within the UK to evaluate children during sleep for evidence of SDB. Based on data from older children, criteria most frequently used to detect abnormality are the mean oxygen saturations (SAT50), the 4% desaturation index (DI4) which is the number of times/hour where the oxygen saturation falls by at least 4% and the delta 12 index (DI12s). This is the average of the absolute differences of saturations (%) between successive 12 s intervals of a pulse oximetry recording resulting in a quantitative measure of the variability in oxygen saturations.11 12

Nocturnal pulse oximetry (NPO) traces are traditionally reported for the entire period of sleep. The differences in the proportion of time spent in REM sleep and the increased susceptibility of infants to central apnoeas mean that reference ranges for desaturation indices and DI12s in older children may not be applicable to young infants. This study aimed for the first time to report reference ranges for oximetry parameters in an infant cohort.


The majority of infants were recruited from the postnatal wards at Princess Anne Hospital, Southampton. The remainder were infants of health professionals who requested inclusion of their new babies in the study. All healthy term infants (37 weeks gestation or more) on the postnatal ward were considered for inclusion. Infants were excluded if they had known respiratory/cardiac conditions or conditions predisposing to SDB. Infants from non-English speaking families were excluded because the study required the ability to understand spoken and written instructions on how to safely use a pulse oximeter.

Written informed consent was taken for both parts of the study during the initial home visit. At this meeting families were advised that if the recording failed a repeat would be offered. Basic demographic and medical history data for the child was obtained from the parents. This included current health status on the date of recording. If the child was unwell the recording was postponed.

Parents were asked to complete a sleep log to ensure accurate analysis of sleep periods.

Data were collected using a Masimo Rad-8 (USA) oximeter with a 2 s averaging time. Adhesive sensors for neonates were placed on either foot. Parents were instructed to keep the probe on throughout the duration of the night’s recording, and to disconnect from the monitor when required, such as during feeds or nappy change. Step-by-step written instructions with images were given to facilitate this. All alarms on the monitor were switched off to limit disturbance to the infant’s sleep.

Families were contacted again when participants approached 3–4 months of age to determine if they were still willing to participate in the follow-up (Recording 2). The method of data collection for Recording 2 was identical to Recording 1.

Data analysis was performed using Visi-Download software (Stowood Scientific, UK). The software automatically extracted artefacts due to low signal identification and quality (IQ), poor perfusion, sensor displacement and motion. Intervals of awakenings were extracted based on information provided by the parents in the sleep log as well as by visual assessment of the trace (eg, rises in pulse rate to presleep levels for an extended duration). Only artefact-free recording time (AFRT) was included in the analysis. Acceptable recordings contained at least 4 hours of data including a period of presumed REM as determined by increased heart rate variability.13 If insufficient data were obtained, a repeat recording was undertaken with the family’s agreement.

Sleep reports were reviewed by a paediatric respiratory consultant with a particular interest in sleep medicine. Based on preliminary data, criteria for elevated clinical suspicion included:

  • mean oxygen saturations below 95%

  • DI4 greater than 30 events per hour

If abnormalities were felt to be technical in nature, a repeat recording was arranged. If the study was felt to be technically acceptable but with clinically suspicious values the child was reviewed and a more detailed cardiorespiratory polygraphy study was offered.

Data were collected for the following indices: mean saturations (SAT50), number of desaturation events >3% (DI3) and >4% (DI4) per hour, minimum saturation percentage (SATmin) and the delta index 12 s (DI12s). Due to the subjective nature of excluding elements of the study by visual assessment, blind inter-rater assessment was undertaken on 10% of studies.

Data analyses were performed using SPSS (V24) and CI analysis software packages.14 Normally distributed variables were reported as means and CI and non-normally distributed variables as medians and CI. For infants that completed both studies paired t-tests and Wilkoxon rank tests were performed to look for outcome differences. To investigate the impact of the length of AFRT on key outcome variables data were analysed after both the first 4 hours and 5 hours of AFRT of Recording 1 and the first 4 hours and 6 hours of AFRT of Recording 2. Paired t-tests were used to assess any differences in group means. In addition, data were collected on the number of infants who switched from being either above or below the 95% CI for NPO variables dependent on the duration of AFRT analysed.


Fifty-six term infants were recruited and underwent NPO at 1 month of age for Study 1 (figure 1). The mean gestational age for these infants was 39+5 weeks (95% CI 39+3 to 40+1). The mean birth weight was 3.37 kg (95% CI 3.24 to 3.41). Less than 4 hours of data were obtained in six infants and the recordings were technically inadequate in seven. Recordings were repeated in three infants and adequate quality data were obtained from two of these. Of the remaining 10 infants 4 were lost to follow-up and it was not possible to arrange a repeat recording in a timely manner in the remaining 6. Acceptable quality data were therefore obtained in 45 (22 boys) infants. The mean age at Recording 1 was 29 days (95% CI 27 to 32). The mean AFRT was 347 min (95% CI 324 to 369). Of the initial cohort 41 infants had repeat overnight oximetry at 3–4 months (Recording 2). Of these, data were insufficient in three cases and it was not possible to arrange repeat recordings for logistical reasons in a timely manner. Mean age at Recording 2 was 112 days (IQR 103–120). Mean AFRT was 418 min (95% CI 385 to 451). Data outcomes are shown in table 1.

Table 1

Mean saturation parameters (95% CIs) for infants studied at 1 month (Recording 1) and 3–4 months (Recording 2) of age providing details of mean oxygen saturations (SAT50), oxygen desaturations > 4% and 3% from baseline (DI4 and DI3), minimum oxygen saturations (SATmin), delta index (DI12s) and percentage time with saturations below 90% and 92%

Figure 1

Recruitment. NPO, nocturnal pulse  oximetry.

Paired data for infants who underwent both studies were available in 32. Analyses demonstrated a significant increase in AFRT (P=0.004), and a significant reduction in DI4 (P=0.006), DI3 (0.03) and DI12s (0.001) (figure 2) and percentage of time with saturations below 90% (P=0.001) and 92% (P=0.000) There was also a significant increase in SATmin between the two groups (P=0.004). An example trace is shown in figure 3.

Figure 2

Scatter plot for paired samples of oxygen desaturations indices >  4 % (DI4) and 3 % (DI3) from baseline and the delta index (DI12s) for Recording 1 and Recording 2.  .

Figure 3

Example of oxygen saturation trace for an infant aged 1 month demonstrating an awake period removed from analysis between 03:00 and just after 04:00, along with clusters of desaturation during periods of REM sleep as evidenced by increased heart rate variability. REM, rapid eye movement.

Blind inter-rater scoring was undertaken for every fifth recording. Reliability coefficients were 0.998 (95% CI 0.984 to 1.000) for mean saturations, 0.994 (95% CI 0.945 to 1.000) for DI4 and 0.995 (0.948 to 0.999) for DI3 indicating excellent inter-rater scoring agreement.

For the two periods of AFRT that were analysed paired t-tests showed no significant differences for group means in DI4, DI3 and DI12s for either Recording 1 (n=29) or Recording 2 (n=27). Using the 95% CI as a cut-off to define normality for individual data resulted in no change of status (‘normal’ versus ‘abnormal’) for any infant in Study 1 for DI4 and DI3. In Study 2 there were changes in status for DI4 in three children and for DI3 in two children dependent on whether the first 4 hours or 6 hours of sleep were analysed. Absolute change in parameters however were small.

One child was felt by the lead clinician to have a clinically significant increase in their desaturation indices between the two recordings. Clinical review was normal, the child went on to have a cardiorespiratory polygraphy study which demonstrated desaturations caused by short central apnoeas and the family declined follow-up.


Oximeters with short averaging times and able to exclude motion artefact are an important advance in the investigation of children for SDB. They are highly sensitive in detecting short changes in oxygen saturation as an indicator of SDB.15 This technology is increasingly being used to wean babies with bronchopulmonary dysplasia out of supplemental oxygen treatment. The objective of this study was to define normal pulse oximetry variables for healthy term infants in the first 4 months of life. Some pulse oximetry parameters using modern oximeters have been described in children over the first 2 years of life.16 However widely quoted criteria for defining abnormality as a result of SDB are the number of falls in oxygen saturations of at least 4% below baseline per hour and desaturations below 90%.4 Normative data have been published in school age children with more than four desaturations greater than 4% from baseline per hour regarded as abnormal.11 In 2012 the American Academy of Sleep Medicine guidelines were amended such that scoring criteria for OSA and CSA were based on a desaturation of 3% from baseline rather than 4%.17 Thus the number of 3% desaturation events per hour may now be considered more relevant than 4%. Little is known about normal limits for 3% desaturations in children and there are no data for young infants.

This study is the first to report both 4% and 3% desaturation indices in young infants. Our data reveal that these indices are much higher than in older children and that these values decrease over the first 4 months of life. This study also reports wide variation for the desaturation indices that were measured. Brockmann et al performed cardiorespiratory polygraphy on infants at 1 month and 3 months of age. They described central apnoea indices at 1 month of age ranging from 0.9 to 44.3 events per hour.18 Since most events in healthy infants are central this natural variability most likely explains the wide oxygen desaturation index range reported in our study.

Our study investigated a community-based cohort. The decision to recruit this study population was based on the fact that most clinical pulse oximetry traces are undertaken at home. The aim was to produce normative data which clinicians could apply to their clinical practice. The limitations of this technique are that interpretation of the sleep-wake state is constrained and dependent on parental reporting. For this reason every fifth trace was independently analysed. This demonstrated excellent agreement in analysis findings.

Our study design limits study duration. Terrill et al reported data on 8–9 hours of infant sleep. In their laboratory-based study participants were continuously supervised by an experienced physiologist able to replace leads if they became dislodged. Our study was a ‘real life’ study dependent on correct parental placement of the oximeter probe after appropriate education. The absolute minimum requirement for a sleep recording is currently unknown. This study focused on a minimum of 4 hours of AFRT including at least one period of REM sleep. We think this is justifiable in young infants who cycle between REM and non-REM sleep every 30 min thus making it easier to capture episodes of active sleep in a shorter time period. This is supported by our cohort of young infants who demonstrated no adjustment of outcomes when data were analysed for either the first 4 hours or 5 hours of sleep. When infants were studied at 3–4 months there were small absolute differences in outcomes for three children dependent on whether data was analysed for the first 4 hours or 6 hours of sleep. AFRT increased significantly between the two time periods and was on average close to 7 hours during Recording 2. The most likely reason for this is that babies were more likely to sleep through the night by 3–4 months and thus there was less likelihood of probe displacement with feeding and crying. A longer minimum period of recorded sleep may be appropriate for older children and requires further study.

This study recruited nine infants of healthcare professionals. In addition all families recruited needed to understand spoken and written English in order to use the equipment safely in the home. These are potential sources of bias. However a subgroup analysis excluding data from infants of healthcare professionals did not significantly change the outcomes suggesting that their inclusion did not materially affect the results. Finally, this study defines variables applicable to motion-resistant oximeters with short averaging times most commonly used to investigate children for SDB. These desaturation indices are not applicable for oximeters set with longer averaging times where desaturations are often blunted by the increased interval between averaging points.19

In conclusion this study is the first to report in infants, reference ranges for oximetry variables routinely used in clinical practice with a motion-resistant oximeter using short averaging times. These data may be of particular use for determining oxygen requirements in young infants. The effect of environmental factors on pulse oximetry values in young infants warrants further study.



  • Funding This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent Not required.

  • Ethics approval Southampton REC (14/SC/0016).

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