Skip to main content

Frequency and duration of extreme hypoxemic and hyperoxemic episodes during manual and automatic oxygen control in preterm infants: a retrospective cohort analysis from randomized studies



Neonatal exposure to episodic hypoxemia and hyperoxemia is highly relevant to outcomes. Our goal was to investigate the differences in the frequency and duration of extreme low and high SpO2 episodes between automated and manual inspired oxygen control.


Post-hoc analysis of a cohort from prospective randomized cross-over studies.


Seven tertiary care neonatal intensive care units.


Fifty-eight very preterm neonates (32 or less weeks PMA) receiving respiratory support and supplemental oxygen participating in an automated versus manual oxygen control cross-over trial.

Main measures

Extreme hypoxemia was defined as a SpO2 < 80%, extreme hyperoxemia as a SpO2 > 98%. Episode duration was categorized as < 5 seconds, between 5 to < 30 seconds, 30 to < 60 seconds, 60 to < 120 seconds, and 120 seconds or longer.


The infants were of a median postmenstrual age of 29 (28-31) weeks, receiving a median FiO2 of 0.28 (0.25-0.32) with mostly receiving non-invasive respiratory support (83%). While most of the episodes were less than 30 seconds, longer episodes had a marked effect on total time exposure to extremes. The time differences in each of the three longest durations episodes (30, 60, and 120 seconds) were significantly less during automated than during manual control (p < 0.001). Nearly two-third of the reduction of total time spent at the extremes between automated and manual control (3.8 to 2.1% for < 80% SpO2 and 3.0 to 1.6% for > 98% SpO2) was seen in the episodes of at least 60 seconds.


This study shows that the majority of episodes preterm infants spent in SpO2 extremes are of short duration regardless of manual or automated control. However, the infrequent longer episodes not only contribute the most to the total exposure, but also their reduction in frequency to the improvement associated with automated control.

Peer Review reports


Continuous of monitoring of peripheral oxygen saturation (SpO2) is the standard of care for preterm infants requiring supplemental oxygenation. Keeping the SpO2 within an acceptable range and minimizing exposure to extremes is associated with better neonatal and long-term outcomes.

Titrating the fraction of inspired oxygen (FiO2) to maintain the SpO2 within a prescribed target range is a challenging task in daily care, resulting a mere 50% compliance within the intended target range [1]. As a result of disordered breathing and apnea, these infants frequently desaturate and caregivers sometimes respond with a transient increase in FiO2 to offset the potential hypoxemia. However, the response with an increase in FiO2 is often slow and inappropriate resulting in extended episodes of hypoxemia as well as swings to hyperoxemia. In addition, a failure to quickly reduce the FiO2 back to baseline levels once the desaturation has resolved further increases the risk of hyperoxemia [2]. The cumulative time in both extreme saturations is impacted by both the frequency and duration of these episodes. Previous studies have shown that the duration in extreme saturations is probably the key determinant for the risk of adverse outcomes [3, 4].

Given the limitations of manual FiO2 titration, automated FiO2 control algorithms for SpO2 have become available and studied. These studies have shown that automated control improves the time within the intended target range, and reduces time spend at the SpO2 extremes [5, 6]. It is unclear how automated FiO2 control impacts the frequency and duration of episodes of extreme hypoxemia and hyperoxemia. Some studies have reported a reduced frequency of one duration (1 min) associated with automated control [5]. More information is needed.

The aim of this post-hoc analysis was to provide additional detailed characterization of the frequency and duration of severe hypoxemic and hyperoxemic episodes, and specifically how they differ between automated and manual FiO2 control.


This was a prospectively defined analysis of retrospective data from randomized cross-over studies investigating automated versus manual FiO2 control in preterm infants 32 weeks gestational age or less in need of invasive or non-invasive respiratory support. The study was carried out in accordance with relevant guidelines and requirements.


This is a post-hoc analysis of a cohort from prospective multicenter randomized cross-over studies [7,8,9,10].


The 7 study sites were tertiary care neonatal intensive care units in 6 different countries.


Eligible subjects who were born less than 33 weeks gestational age were considered from four cross-over studies investigating manual versus automated FiO2 control (AVEA-CLiO2, Vyaire Mettawa IL, USA) in which individual subject data was available [7,8,9,10]. These studies used either two 24 or 12 hour intervention periods. To avoid potential bias, the two selection criteria were made prospectively, that is, without reviewing the outcomes. Since these individual studies used different SpO2 targets, and we did not want to confound the results, we selected only studies using a SpO2 mid-point of 90% and a target range width of 4%. This excluded two studies [9, 10]. Finally, subjects in the two remaining studies [7, 8] that did not spend at least 75% of time on supplemental oxygen were excluded to permit better characterization of hyperoxemic episodes. All these cases are from clinical trials with ethics approval, and are de-identified. All subjects were enrolled with written informed consent of their guardian(s).

Outcome measures

We defined the primary SpO2 exposure metrics for SpO2 extremes as < 80% (hypoxemia) and > 98% (hyperoxemia). Episodes of 600 seconds or longer were excluded as being likely related to procedures and not representative of routine SpO2 management. The primary endpoint was the percent time of SpO2 episodes in each of 5 episode-duration categories (< 5 s, 5 to < 30s, 30 to < 60s, 60 to < 120 s, 120 to < 600 s). We also determined the frequency of episodes within each duration epoch. This categorization was based on data collected with a resolution of every 5 seconds, and two consecutive data points were defined as 5 seconds.

Statistical analyses

With the sample of 58 subjects, we determined that we would have a > 80% chance of detecting an absolute difference of 10%-time, assuming an absolute variance of 25%-time, with an uncertainty of p < 0.05.

Extraction of the endpoints from the 5-second data points in the database was accomplished with purpose-built software (MatLab, Mathworks, Natick MA USA). Differences among the episode length categories and mode of FiO2 control were determined with the Kruskal-Wallis test with Dunn’s procedure for pairwise comparisons. A two-tailed p < 0.05 was considered statistically significant. Statistical tests were conducted with XLSTAT v19.03 software (Addinsoft, Paris, France).


From the initial 179 potential subjects, using the prospective selection criteria, we made the following exclusions: two studies (59 subjects) were excluded because of their target range [9, 10], 40 subjects were excluded from one study that evaluated high and low target range cohorts [7] and finally, due to higher exposure to room air, in the remaining two studies [7, 8], 22 subjects were excluded (12, 10 respectively). Thus, we evaluated the SpO2 control of 58 preterm infants receiving respiratory support and supplemental oxygen. Fifty-one percent of the cases came from one study [7], and the rest from the second study [8]. The inspired oxygen in these studies was controlled manually (M-FiO2) for 1 day and automatically (A-FiO2) on the other, in random order. All the infants were on the same mode of respiratory support (noninvasive or intubated) during the days of manual and automated control. Most were managed noninvasively (83%). The demographics of the subjects are shown in Table 1, they were mostly extremely preterm and were studied weeks after birth. The subjects’ baseline oxygen needs were relatively low (median FiO2 0.28, IQR 0.25-0.32). A majority of the subjects spent no time on room air (59%). Those who did have periods without supplemental oxygen, spent a nominal amount of time without (median 4%, IQR 1-9%). The median SpO2 for both the automated and manual control methods were nearly identical (91%). However, there was a difference between the groups in the time spend outside normoxemia; the SpO2 was < 87% for 13% of the time during A-FiO2, and 16% during M-FiO2, whereas 11% of the time the SpO2 was > 95% during A-FiO2, and 16% during M-FiO2. These differences were statistically significant (p < 0.001).

Table 1 Subject demographics

There were frequent episodes at SpO2 extremes with differences between the two methods of control as shown in the Table 2. Episodes at the SpO2 extremes (< 80 and > 98%), tended to be less frequent during A-FiO2 within each of the 5 episode-duration categories. Most of these extreme episodes were shorter than 30 seconds.

Table 2 Frequency of extreme hypoxemic and hyperoxemic episodes

Our primary endpoint is shown in the Fig. 1. It presents the total time at< 80% and at > 98% SpO2, along with the contribution towards the total from each of the 5 episode-duration categories. The difference in the total time for A-FiO2 and M-FiO2 at each of these two endpoints was significantly different (p < 0.001), favoring A-FiO2. The time at each of the episode length categories tended to be lower during A-FiO2, and nearly all differences were statistically significantly significant [all but 5 to < 30 seconds with SpO2 < 80%]. Both shorter and longer episodes made a relevant contribution to the total duration of extreme SpO2 exposure with about half of the duration of time in hypoxemia < 80% resulted from episodes with a duration of 30 seconds or shorter, whereas one third of the duration of time with hyperoxemia > 98% was a result of episodes 30 seconds or shorter. In contrast, the duration of time < 80% from episodes 60 seconds or longer were 19% for A-FiO2 and 49% for M-FiO2. More than a third of the duration of time > 98% were from episodes 60 seconds or longer (34% A-FiO2, 38% M-FiO2). Importantly, though infrequent during both modes of control, nearly two-thirds of the reduction of total time spent at the extremes between automated and manual control (3.8 to 2.1% for < 80% SpO2 and 3.0 to 1.6% for > 98% SpO2) was from a reduction of episodes of 1 min or longer.

Fig. 1
figure 1

Stacked Histogram of % Time at SpO2 Extremes. Differences in % Time between automated fraction of inspired oxygen (A-FiO2) control and manual (M-FiO2) control: Total %Time (A-FiO2-M-FiO2):> 98% p < 0.001, < 80% < 0.0O1, For each of the 5 duration categories (ns, * < 0.05, ** < 0.01, *** < 0.001)


In a population cohort of extremely preterm infants receiving supplemental oxygen and respiratory support, we investigated the frequency and duration of marked hypoxemic and hyperoxemic episodes during both A-FiO2 and M-FiO2. We confirmed that the infrequent longer episodes were the primary contributor to total time at SpO2 extremes. We found that, compared to M-FiO2, automated control of SpO2 reduced the number of episodes in all durations, but that most of the reduction in the total exposure to these extremes was from a reduction of episodes of 1 min or longer.

Though we provide more detail, our hypoxemic episode results are consistent with other studies. The extensive report of Poet et al., found that hypoxemic exposure with manual FiO2 control is comprised of episodes of desaturations across a range of durations, with prolonged episodes being infrequent, but nevertheless contributing markedly to the total exposure to SpO2 less than 80% [3]. Other studies using this automated FiO2 control algorithm have reported a reduced frequency of hypoxemic episodes longer than a minute, compared to manual FiO2 control [9, 11]. Our study identified that the duration of hypoxemic episodes is the primary factor in A-FiO2 reducing total time in hypoxemia. One study of A-FiO2 reported that the total number of shorter episodes below the target range was increased but that the longer episodes were decreased with A-FiO2, suggesting that potentially longer episodes were compressed [9]. This was not confirmed in our study. Rather we found that automated FiO2 control also reduced the frequency of shorter episodes of hypoxemia. We suggest that this different finding is a result of a higher median SpO2 during manual in the previous report. The impact of such a shift is consistent with another study [7].

We believe our data are the first report of the details of the frequency and duration of episodes of hyperoxemia. We found, consistent with desaturations, that it is comprised of episodes across a range of duration, with prolonged episodes being infrequent, but nevertheless contributing markedly to the total exposure to SpO2 greater than 98%. Studies of this automated FiO2 control system have also shown a reduction of the frequency of hyperoxemic episodes longer than a minute [8,9,10,11], but our study demonstrated that this is the primary factor in reducing total time in hyperoxemia with A-FiO2.

There is limited information on how these differences might impact outcome, that has come from rigorous post-hoc analyses of the large randomized trials. One report found that better control of the SpO2, regardless of the target range, improved long term outcomes but they did not report the associated exposure to SpO2 extremes [12]. Another reported an association between cumulative exposure to SpO2 < 80%, which correlated to the number of prolonged episodes, with an increased risk of late death or disability at 18 months and also speculated that therapies that reduced these prolonged events could improve long-term outcomes [3]. Another research team reported that an increased frequency of all hypoxemic events was associated with severe retinopathy of prematurity and bronchopulmonary dysplasia [13, 14]. We speculate that more frequent desaturations would also correlate with more prolonged hypoxemic episodes, as well as an increased risk of overshoot to hyperoxemia. The impact of hyperoxemia has not been so carefully evaluated, but a landmark study published nearly 20 years ago confirmed that high levels of SpO2 are associated with severe retinopathy of prematurity and bronchopulmonary dysplasia, without a difference in hypoxemia [4]. Regardless, it is not clear whether these studies report a cause-effect relationship, or rather a marker of adverse outcome.

Alarm setting strategies ought to be studied much more. They should strive to reduce the time at SpO2 extremes, by specifically balancing the risk of missing important events because of false negatives versus not responding to important events due to alarm fatigue. These results may help in formulating strategies to reduce oximeter alarm fatigue by considering the trade-off between false-positive and false-negative alarms that is associated with the alarm delay. Our data show that during manual FiO2 control, most of the episodes of extreme SpO2 levels in this group of infants resolved on their own in less than 30 seconds. One might question the clinical relevance of the settings that trigger highly frequent alarms not needing attention. Nevertheless, these frequent alarms might be useful in alerting nurses to instability. In contrast, it seems reasonable that an A-FiO2 control system, which continuously makes FiO2 adjustments as often as many times per minute, might need different alarm delays. It is important to note that in the case of A-FiO2 control, a saturation alarm indicates that changes in the FiO2 have not adequately mitigated the alarm condition, and thus personal attention is needed. Our data suggest that alarms delay during A-FiO2 might be set at 60 seconds or longer in order to reduce false alarms. One study supports these considerations [15].

We evaluated one A-FiO2 system and the findings of this study should be generalized to other A-FiO2 systems cautiously. Alternative algorithms, that consider changes much less frequently than every second would not be expected to reduce episodes of shorter duration. Nevertheless, we also found that much of the reduction of total exposure was a result of reducing the duration of longer episodes, so this might not be clinically relevant. In contrast more frequent automatic adjustments might lead to more overshoot between extremes, which is a constant concern during manual control. Overshoot using the A-FiO2 algorithm that we studied has shown to be better than M-FiO2 [9, 16], but this needs to be evaluated in other A-FiO2 systems, regardless of their frequency of adjustment.

The primary limitation of our study is that it reflects a small population of infants and a single day of exposure. However, the results do seem to be consistent with other studies. The thresholds for extreme levels of SpO2 were prospectively defined, and have been shown to be associated with extreme levels of PaO2 [17]. Nevertheless, evaluation of other, less extreme, SpO2 thresholds might yield different conclusions and also impact outcomes. Nevertheless in a post hoc analysis we found the effects of using cut offs of < 87% and > 95% were similar, suggesting the findings would be insensitive to the exact cut off. Further we counted episodes > 98% even if the subject was temporarily not receiving supplemental oxygen. This would tend to increase the frequency of such episodes, but our analysis population, had a very limited amount of time without supplemental oxygen. There are also some other aspects of the study impacting the generalization of our work. First, our study reflects experience with a lower SpO2 target range resulting in a median SpO2 of 91%. A higher target range, if resulting in a higher median SpO2, would likely have a different profile. One study comparing two ranges (89-93% and 91-95% SpO2) found a shift to less hypoxemia and more hyperoxemia during both automated and manual FiO2control [8]. Studies also suggest that M-FiO2 control of SpO2 at higher target ranges is easier to manage than at lower target ranges [7, 18]. Finally, this study population was not large enough to explore relative differences associated with invasive and noninvasive ventilation or during periods of differing stability.


Our study characterizes the distribution of extreme hypoxemic and hyperoxemic episodes. Nearly all of these extreme episodes are of short duration. However, the relatively infrequent longer episodes were the main contributors to the total SpO2 extreme exposure. We demonstrated that A-FiO2 results in fewer episodes of all durations, but that a reduction in the infrequent longer episodes was the main factor in its improved effectiveness. We believe this information might be useful not only in refining SpO2 alarm practices, but also for refining and comparing the effectiveness of A-FiO2 control systems.

Availability of data and materials

The data sets analyzed in this study are not publicly available, as a result of the investigator agreements when conducted, but are available from the corresponding author upon reasonable request.


SpO2 :

Peripheral oxygen saturation

FiO2 :

Fraction of inspired oxygen


Interquartile range


Automated control of FiO2


Manual control of FiO2


  1. Hagadorn JJ, Sink DW, Buus-Frank ME, et al. Alarm safety and oxygen saturation targets in the Vermont Oxford Network iNICQ 2015 collaborative. J Perinatol. 2017;37:270–6.

    Article  CAS  Google Scholar 

  2. van Zanten HA, Tan RN, Thio M, et al. The risk for hyperoxaemia after apnoea, bradycardia and hypoxaemia in preterm infants. Arch Dis Child Fetal Neonatal Ed. 2014;99(4):F269–73.

    Article  Google Scholar 

  3. Poets C, Roberts R, Schmidt B, et al. Association between intermittent hypoxemia or bradycardia and late death or disability in extremely preterm infants. JAMA. 2015;314(6):595–603.

    Article  CAS  Google Scholar 

  4. Askie LM, Henderson-Smart DJ, Irwig L, Simpson JM. Oxygen-saturation targets and outcomes in extremely preterm infants. NEJM. 2003;349:959–67.

    Article  CAS  PubMed  Google Scholar 

  5. Mitra S, Singh B, El-Naggar W, McMillan DD. Automated versus manual control of inspired oxygen to target oxygen saturation in preterm infants: a systematic review and meta-analysis. J Perinatol. 2018;38(4):351–60.

    Article  Google Scholar 

  6. Abdo M, Hanbal A, Asla MM, et al. Automated versus manual oxygen control in preterm infants receiving respiratory support: a systematic review and meta-analysis. J Matern Fetal Neonatal Med. 2021:1–8.

  7. van Kaam AH, Hummler H, Wilinska M, et al. Automated versus manual oxygen control with different saturation targets and modes of respiratory support in preterm infants. J Pediatr. 2015;167(3):545–50.

    Article  Google Scholar 

  8. van den Heuvel ME, van Zanten HA, Bachman T, et al. Optimal target range of closed-loop inspired oxygen support in preterm infants: a randomized controlled study. J Pediatr. 2018;197:36–41.

    Article  Google Scholar 

  9. Claure N, Bancalari E, D’Ugard C, et al. Multicenter crossover study of automated control of inspired oxygen in ventilated preterm infants. Pediatrics. 2011;127(1):e76–83.

    Article  Google Scholar 

  10. Lal M, Tin W, Sinha S. Automated control of inspired oxygen in ventilated preterm infants: crossover physiological study. Acta Paediatr. 2015;104(11):1084–9.

    Article  CAS  Google Scholar 

  11. Waitz M, Schmid MB, Fuchs H, et al. Effects of automated adjustment of the inspired oxygen on fluctuations of arterial and regional cerebral tissue oxygenation in preterm infants with frequent desaturations. J Pediatr. 2015;166(2):240–4.

    Article  Google Scholar 

  12. Schmidt B, Whyte RK, Shah PS, et al. Effects of targeting higher or lower oxygen saturations in centers with more versus less separation between median saturations. J Pediatr. 2016;178:288–91.

    Article  Google Scholar 

  13. Raffay TM, Dylag AM, Sattar A, et al. Neonatal intermittent hypoxemia events are associated with diagnosis of bronchopulmonary dysplasia at 36 weeks postmenstrual age. Pediatr Res. 2019;85(3):318–23.

    Article  Google Scholar 

  14. Di Fiore J, Bloom J, Orge F, et al. A higher incidence of intermittent hypoxemic episodes is associated with severe retinopathy of prematurity. J Pediatr. 2010;157(1):69–73.

    Article  Google Scholar 

  15. Warakomska M, Bachman TE, Wilinska M. Evaluation of two SpO2 alarm strategies during automated FiO2 control in the NICU: a randomized crossover study. BMC Pediatr. 2019;19(1):142.

    Article  Google Scholar 

  16. Bachman TE. Prevalence of potentially clinically relevant complex episodes of extreme SpO2 during manual and automatic control of inspired oxygen. Lek Tech. 2022;52(1):84–90.

    Google Scholar 

  17. Bachman TE, Newth CJL, Iyer NP, et al. Hypoxemic and hyperoxemic likelihood in pulse oximetry ranges: NICU observational study. Arch Dis Child Fetal Neonatal Ed. 2019;104(3):F274–9.

    Article  Google Scholar 

  18. Stenson BJ. Achieved oxygenation saturations and outcome in extremely preterm infants. Clin Perinatol. 2019;46(3):601–10.

    Article  Google Scholar 

Download references


This analysis and manuscript was not funded.

Author information

Authors and Affiliations



TB and KR conceptualized the study and integrated the database. TB, WO and AvK were major contributors to the writing of the manuscript. WO, AvK, HH, ML and CF managed the testing of the subjects from their center. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Thomas E. Bachman.

Ethics declarations

Ethics approval and consent to participate

The data used in this analysis is a subset of data from two published multicenter studies. Cases were from 7 hospitals, each received approval to participate, including a requirement for written informed consent, by the authorized ethics authority. It was collected at the following hospitals, the authorizing ethics authority for each site is noted. The centers, in order of number of subjects included, are as follows: Emma Children’s Hospital, The Netherlands [AMC University of Amsterdam]; University Medical Center of Ulm [University of Ulm]; Leiden University Medical Center, The Netherlands [AMC University of Amsterdam]; Silesian Institute Mother and Newborn, Poland {Medical University of Silesia]; James Cook University Hospital, United Kingdom [Health Research Authority, Newcastle & Tyneside]; Vittore Buzzi Children’s Hospital, Italy [University of Milan]; Alberta Children’s Hospital, Canada [University of Calgary] All experimental protocols were approved by the named institutional and/or licensing committee.

Consent for publication

Not applicable, the manuscript contains no person or copyrighted material.

Competing interests

The institutions of TB, WO and AvK receive funding from the manufacture of the ventilator used in this study, but that funding did not include the work of the study. GL receives educational honoraria from the manufacture of the ventilator. The other authors declare no competing interests. The manufacture of the ventilator used in this study had no part in the study design and has not reviewed the results.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Bachman, T.E., Onland, W., van Kaam, A.H. et al. Frequency and duration of extreme hypoxemic and hyperoxemic episodes during manual and automatic oxygen control in preterm infants: a retrospective cohort analysis from randomized studies. BMC Pediatr 22, 350 (2022).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: