Objective To assess the efficacy of a newly developed system for closed loop control of the fraction of inspired oxygen (FiO2) on variation of arterial (SpO2) and on regional tissue oxygen saturation (StO2) in preterm infants with fluctuations in SpO2.
Design Randomised crossover trial comparing automated (auto) to manual FiO2 adjustment (manual) during two consecutive 24 hours periods using a Sophie infant ventilator (SPO2C).
Setting Tertiary university medical centre.
Patients Twelve very low birthweight infant (VLBWI) (gestational age (median; IQR): (25; 23–26 weeks); birth weight (mean±SD): (667±134 g); postnatal age (mean±SD): (31.5±14 days)).
Main outcome measure Time within SpO2 target range.
Results There was an increase in time within the intended SpO2 target range (88%–96%) during auto as compared with manual mode (77.8%±7.1% vs 68.5%±7.7% (mean±SD), p<0.001) and a decrease in time below the SpO2 target during the auto period (18.1%±6.4% vs 25.6%±7.6%; p<0.01). There was a dramatic reduction in events with an SpO2 <88% with >180 s duration: (2 (0–10) vs 10 (0–37) events, p<0.001) and the need for manual adjustments. The time the infants spent above the intended arterial oxygen range (4.1%±3.8% vs 5.9%±3.6%), median FiO2, mean SpO2 over time and StO2 in the brain, liver and kidney did not differ significantly between the two periods.
Conclusions Closed-loop FiO2 using SPO2C significantly increased time of arterial SpO2 within the intended range in VLBWI and decreased the need for manual adjustments when compared with the routine adjustment by staff members. StO2 was not significantly affected by the mode of oxygen control.
- closed loop, oxygen, BPD
Statistics from Altmetric.com
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?
Very low birth weight infant (VLBWI) change their respiratory effort often very rapidly resulting in hypoventilation or apnoea, or may expire actively by increasing intrathoracic pressure resulting in loss of functional residual capacity and hypoxaemia.
Hypoxic episodes may have a clinically relevant impact on overall health conditions in preterm infants.
Several investigators have shown that automated fraction of inspired oxygen (FiO2) adjustments may limit the fluctuations in arterial oxygen saturation (SpO2) in preterm infants with frequent desaturations.
What this study adds?
Closed-loop FiO2 using a newly developed FiO2 closed-loop system significantly increased time of arterial SpO2 within the intended range in VLBWI when compared with the routine adjustment by staff members.
Oxygen tissue saturation was not significantly affected by the mode of oxygen control.
Many very low birthweight infants (VLBWI) require respiratory support including supplemental oxygen and/or positive pressure during the first days or weeks of life. VLBWI change their respiratory effort often very rapidly resulting in hypoventilation or apnoea, or may expire actively by increasing intrathoracic pressure resulting in loss of functional residual capacity (FRC) and hypoxaemia.1 2 Depending on the frequency, duration and severity, these hypoxaemic episodes may have the potential to cause brain or other organ damage.3 The clinical staff taking care of these infants often increases the fraction of inspired oxygen (FiO2) to maintain arterial oxygen saturation as measured by pulse oximetry (SpO2) within the intended target range. Once recovering from these hypoxaemic episodes, these FiO2 adjustments often result in SpO2 above the target range,4 which may predispose the infants for hyperoxic tissue damage of the brain and lung,3 5 or for retinopathy of prematurity.6 Cerebral oxygen desaturation may predict cognitive impairment in adults undergoing surgery,7 and evidence is now available to show that hypoxic episodes may have a clinically relevant impact on overall health conditions in preterm infants.3 Furthermore, in infants with frequent desaturations it is often very difficult and time consuming for the staff to maintain the infants’ SpO2 within the intended target range.8 Several investigators have shown that automated FiO2 adjustments may limit the fluctuations in SpO2 in preterm infants with frequent desaturations.9–13 Waitz et al showed a significant increase of the time within the SpO2 target range and a reduced incidence of prolonged hypoxaemic events compared with manual FiO2 adjustment.13 In two randomised crossover multicentre trials improved maintenance of the intended SpO2 range and a reduced time with SpO2 above the target range was found, when infants were exposed for 24 hours to automated FiO2 adjustment as compared with manual control.14 15 Desaturations often coincide with bradycardia, which may decrease cardiac output, impair tissue perfusion and therefore result in cerebral hypoxia. There is very little data available on the effect of automated FiO2 adjustments on tissue oxygenation of different organs in preterm infants.13 In this study, we have tested the hypothesis that in preterm infants with a postmenstrual age of <30 weeks gestational age with frequent desaturations requiring respiratory support, automated FiO2 adjustment using a newly developed closed loop mode based on measured SpO2 in comparison with manual FiO2 adjustment improves the time of arterial oxygen saturation as well as tissue oxygenation of different organs within the target range during a 24-hour experimental time.
This study was a randomised crossover clinical trial comparing two treatment phases of 24 hours duration each, clinical (manual) routine and automated adjustment of FiO2 (auto). Study infants were recruited in the neonatal intensive care unit (NICU) of the children’s hospital, Ulm University Medical Center. Approval of the study was obtained by the institutional review board of the University of Ulm prior to recruitment of patients. Preterm infants with frequent desaturations requiring non-invasive or invasive ventilator support were eligible for this study if they fulfilled the following criteria: gestational age <30 weeks, at least four desaturations (SpO2 <80%) during an 8-hour period within the 24 hours before the study using a standard pulse oximeter of our NICU (Masimo Radical 7) and informed consent obtained from the parents or legal guardian. Exclusion criteria were: major congenital anomalies, clinical evidence for seizures or ongoing sepsis (C reactive protein >10 mg/L or positive blood culture or requirement of catecholamine infusion), congenital cyanotic heart disease, no decision for full treatment support, average FiO2 during the last 24 hours before the active study phase >0.60 or expected need of blood transfusion during study period. Furthermore, we did not study infants within the first 96 hours of life to exclude rapidly changing conditions during the early phase of respiratory distress syndrome and to avoid additional handling of the infant during the critical period for intraventricular haemorrhage.
Randomisation of the sequence of the two study phases was carried out using sealed envelopes. Infants were changed to a newly developed specific ventilator device approved for clinical use in neonates in Europe, which is capable to automatically adjust the FiO2 based on readings of a connected SpO2 monitoring device (Sophie Respirator, Fritz Stephan GmbH Medizintechnik GmbH, Gackenbach, Germany), and which has an integrated automated oxygen control (‘SPO2C’) using standard pulsoximetry technique: Radical (Masimo, Irvine, California, USA, averaging time 8 s).
Technically, the ‘SPO2C’ is a dual PID controller. ‘P’ stands for ‘Proportional,’ ‘I’ for ‘Integral’ and ‘D’ for ‘Differential.’ The ‘P’ component looks at the difference between the measured and target values. The ‘I’ component includes the change of the difference over time, while the ‘D’ component addresses the speed of changes. The primary, innermost control circuit has variable control coefficients. The used coefficients depend on the range into which the controlled variable (ie, the measured SpO2) falls. In addition, parameters are adjusted depending on whether saturation is increasing or decreasing. The arithmetic mean of the set SpO2 target range serves as the reference (set value). The FiO2 set value represents the actuating variable, and is controlled using an adjustable base FiO2 value. This means that even in the event of a consistently insignificant control deviation, the actuating variable does not adjust to the minimum possible FiO2 (0.21). Instead, the base FiO2 is dispensed. The control parameters are selected in such a way that the controller can quickly respond to decreases of SpO2 below the target range to prevent long episodes of severe desaturations. The controller responds to variations in SpO2 above the lower limit of the target range with moderate adjustments of FiO2. This will attenuate FiO2 fluctuations when SpO2 falls within the target range. The controller’s differential component is amplified below this target range. This enables quick responses to significant desaturations. The controller’s proportional components are exponentially weighted depending on the set base FiO2 in all ranges. This links the oxygen binding curve to the controller. This curve is not linear and expresses the fact that at high oxygen saturation or at a higher FiO2 greater differences in FiO2 are needed to achieve changes to SpO2.
A second control circuit adapts the base FiO2. The controller is set so that changes in SpO2 over a period of 5 min lead to an indirect change of the base value. The controller can be disabled by various events for safety reasons. In this case, the ‘Backup FiO2’ setting is used for ventilation. After the ‘SPO2C’ is switched on, the FiO2 is adjusted by the value of the set base FiO2 based on the measured SpO2 and on the target range at intervals of 2 s. The increment of change is determined by the difference between the measured SpO2 and the SpO2 target value, as well as the rate of change of the SpO2. The greater the difference between the measured SpO2 and the SpO2 target value, the greater the change in FiO2. The faster the change in measured SpO2, the greater the change in FiO2. If the measured SpO2 is outside of the target range, the degree of decrease or increase also becomes greater.
The ‘fluctuations’ in FiO2 are captured and used to correct the base FiO2 by the second control circuit in the following way: If the average FiO2 delivered is lower than the base FiO2, the base FiO2 is decreased and if the average FiO2 delivered is higher than the base FiO2, the base FiO2 is increased. This correction of the base FiO2 value occurs every 5 min. Technically, this correction also represents a PID controller. By adjusting the base FiO2, the controller minimises the deviation between the mean FiO2 and the base FiO2.
After randomisation the infants were exposed to the first study phase (clinical routine or automated FiO2 adjustment) for 24 hours and then switched to the alternate mode for another 24 hours. The SpO2 monitor connected to the ventilator was used for clinical SpO2 monitoring during the study. The SpO2 sensor was attached to the right arm. An additional sensor was attached to the infants’ forehead for continuous measurement of cerebral, to the flank for renal and to the right abdomen for hepatic tissue oxygen saturation monitoring using a near-infrared spectroscopy device (Noonin device SenSmart X-100 Oximeter System). SpO2 values, FiO2, ventilator settings and heart rate as determined by ECG (GE Dash 4000, Freiburg, Germany) were recorded every 2 s via electronic data logger. The selected target range for preductal SpO2 was 88%–96% and alarm limits for interventions were selected at 87% (lower limit) and 97% (upper limit) during both ventilator conditions.
The infants were receiving continuing care in the NICU as judged necessary by the clinical team. However, the clinicians were instructed to avoid the following interventions which may influence the variables of interest unless this was judged necessary for clinical care: muscle paralysis, changes in management with drugs for sedation, analgesia or anticonvulsive medication, changes in caffeine or theophylline dosing or changes in continuous positive airway pressure (CPAP) settings during the active study period. Binasal CPAP prongs were used during the study period in study subjects on non-invasive ventilation. Infants with sudden deterioration requiring rescue treatment such as high frequency ventilation and/or inhaled nitric oxide were to be excluded from the study but to be reported.
Protocol for FiO2 adjustments: the physician and nursing staff were encouraged to follow the following guidelines on FiO2 adjustments: if SpO2 was above target range, reduction of the FiO2 by 0.02–0.05 steps every 2–5 min was attempted. If SpO2 was below target range, infant assessment and an increase in FiO2 by 0.05–0.20 manually or by using the ‘increase O2-button’ (preset to +0.20) was suggested. FiO2 was adjusted every 30–120 s according to the clinical response. FiO2 was not increased if SpO2 was already trending up. If SpO2 was below 80% or the infant had bradycardia (HR <80/min), the infant was assessed and FiO2 was increased by 0.20–0.40. Furthermore, the infant was assessed until SpO2 returned to the assigned range and FiO2 was set to baseline when this occurred. The clinical attending was informed if there was a persistent increase in baseline FiO2 of >0.20. The study investigator was notified immediately if any malfunction of the device was suspected.
Sample size and statistics
The primary outcome measure was the total duration of time with a SpO2 (measured as the percentage of the total recording time) within the target range (88%–96%). Secondary outcomes included the total time (proportion of the total recording time) with SpO2 <80%, <70%, the time with SpO2 >96%, the number of episodes below the SpO2 target range (defined as episodes with a duration of >10 s/60 s/180 s), mean SpO2, the number of bradycardia events (defined as a heart rate <100/min for >10 s duration and the mean FiO2 during each period of the study time.
For the detection of fluctuations in cerebral, renal, hepatic tissue oxygenation, the area under the curve (AUC) (%×s) below and above the individual median cerebral, hepatic and renal StO2 range (defined as ±5% of the individual StO2 median) was assessed for each infant during each period as described before.13 Furthermore, StO2 of brain, liver and kidney was evaluated as the median cerebral, hepatic, renal StO2, variability of cerebral, hepatic, renal StO2, the proportion of time with cerebral, hepatic, renal StO2 <55%, and the time (as percentage of the total recording time) and the number of events (duration >2 s) with StO2 below and above the defined cerebral StO2 range.
Due to the crossover design with exposure of the same patients to the two different ventilatory modes, we compared paired data. The sample size calculation was based on the assumption that there were no carry-over effects. The primary outcome variable was the time with a SpO2 within the target range (88%–96%). In another study, we had measured the effect of using different SpO2 target ranges during manual FiO2 control on the actual time the infants remained within that range.16 According to the data obtained, the difference in time with arterial oxygen saturation within the two target ranges studied was 71.6±56.3 min (mean±SD). Our sample size calculation was based on the assumption that the intervention (SPO2C) was at least as effective to improve the time of SpO2 within the target range as using the higher SpO2 range in the comparison of the intervention mentioned above. Based on an α-error of 0.05 and a β-error of 0.10 (power of 90%), a sample size of nine patients was needed to be able to show the same difference. To adjust for a potential loss of data (defined as a signal loss of >30% during the study time), the decision was to increase the sample size by 30% and to study 12 patients. All data were continuous variables and for statistical comparisons we used paired t-tests or Wilcoxon signed-rank tests where appropriate. The level of significance was a p<0.05 for the primary outcome. All other measures were compared in a descriptive way only (thus hypothesis generating) and the level of significance for these comparisons was a p<0.05.
A total of 12 preterm infants with intermittent hypoxaemia were included from March 2014 to September 2014. The median (range) gestational age was 25 weeks (23–26) with a birth weight of 667 g (490–880). The median age at study entry was 31 days (12–62) and weight was 1009 g (600–1700). Ten infants were on non-invasive ventilation mode with CPAP or non-invasive intermittent positive airway pressure and two infants were on invasive ventilation (table 1).
The time within the SpO2 target range (88%–96%) was significant increase during the automated FiO2 control as compared with the manual period (77.8%±7.1% vs 68.5%±7.7% (mean±SD), p<0.001, table 2). Significant less time below the SpO2 threshold during the automated as compared with the manual period (18.1%±6.4% vs 25.6%±7.6%) was detected (p<0.01). There was a significant reduction in events with an SpO2 <88% with >60 s duration and dramatic reduction in these events with >180 s duration (table 2). Mean SpO2 was significantly higher during the automated mode, although the difference was very small (table 2). The percentage of time the infants spent above the intended arterial oxygen target range and mean FiO2 over time did not differ significantly between the two periods (table 2). The number of bradycardia events during both periods did not vary significantly between the two periods (7.2 (0–18) vs 6.9 (0–22)) (table 2). The median number of manual adjustments of the FiO2 was significantly reduced during manual as compared with automated period (adjustments per hour 7.5 (2.1–14.3) vs 0.5 (0–1) (table 2).
The total time below/above the individual cerebral StO2 threshold as defined in the ’Methods' section, the median cerebral StO2, the time with a cerebral StO2<55% or the time below or above the defined threshold as well as the total number of events below or above the defined cerebral StO2 thresholds was not significantly different comparing both periods (table 3).
The total time below/above the individual hepatic/renal StO2 threshold, the median hepatic/renal StO2, the time with a hepatic/renal StO2<55% or the time below or above the thresholds as well as the total number of events below or above the defined hepatic/renal StO2 thresholds was not significantly different comparing both periods (tables 4 and 5).
In this study, we evaluated the effect of automated FiO2 adjustment as compared with routine manual FiO2 control using a novel device (SPO2C) on SpO2 and cerebral, hepatic, and renal StO2 in preterm infants with frequent SpO2 fluctuations. The results of our study showed a significant improvement in time within intended SpO2 target range during the automated FiO2 control. The increased time within the intended arterial oxygen saturation range was accomplished by a significant reduction in the proportion of time with hypoxaemia and a trend towards less time with hyperoxaemia during the automated oxygen supplementation.
The efficacy of SpO2 targeting, manually or automated, is affected by the infant’s overall respiratory stability. All infants had at least four desaturations with an SpO2<80% during 8 hours prior to the study. The number of events SpO2<80%/24 hour, median (range) prior to the study phase was 35 (5–59). Frequent hypoxaemic events in preterm infants especially very long episodes, have been shown to be associated with significant side effects,3 including severe retinopathy of prematurity.6 More recently, published data from the Canadian Oxygen Trial suggests that frequent prolonged hypoxaemic episodes (>60 s) in extremely preterm infants during the first 2–3 months after birth may be associated with an increased risk of late death or disability at 18 months.17 In our study, the frequency of prolonged events with SpO2<88% with a duration of >60 s was reduced during the automated period. Furthermore, the effect was even more pronounced for events lasting >180 s. We have also evaluated the daily frequency of episodes with SpO2<80% with>10 s and >60 s duration. We found a significant reduction of longer events (events with SpO2<80% (>60 s)/24 hours, median (range) manual 75 (22–165) versus auto 43 (3–60) p<0.001 and events SpO2<80% (>180 s)/24 hours, median (range) manual 6 (0–25) versus auto 1 (0–2), p<0.001).
However, currently there are no data from randomised controlled trials on long-term use of automated FiO2 control available to show that its use is associated with decreased mortality or morbidity, such as retinopathy of prematurity or impaired neurological outcome in preterm infants.
The secondary aim of our study was to evaluate the effect of automated FiO2 adjustment on fluctuations of tissue oxygenation. As the normal range of StO2 is not well defined for preterm infants and we were particularly interested in fluctuations, we defined an individual StO2 reference range for each infant during each period, similar as in the study by Waitz et al.13 This StO2 reference range was arbitrarily defined as the individual median cerebral StO2±5%. Currently, there is limited information what is the best StO2 target range for our study infants and we actually do not know if a certain StO2 is the ’right’ threshold for intervention. Furthermore, there is no single StO2 corresponding to a specific SpO2 value, as StO2 is influenced by perfusion and oxygen extraction, that is, oxygen consumption that can change dramatically over time, especially in the brain.
We did not find a significant difference in the AUC (%×s) above or below the cerebral StO2 reference range, suggesting that there is no significant difference between the two modes of FiO2 control. The number of events and the proportion of time above or below the cerebral StO2 range were also similar during both periods. We can speculate that automated FiO2 control was not able to further improve stability of StO2 in the different organs studied. However, our infants were studied at a median age of more than 1 month, and infants at the age of <96 hours were not eligible for this study. Autoregulation of organ perfusion including the brain and other organs may be already sufficiently developed at this postnatal age in preterm infants. We speculate that the findings may be different when studying very immature infants during the first week of life. However, on the other hand we cannot fully exclude a methodological artefact in StO2 data analysis. However, even if automated FiO2 control does not improve fluctuations of StO2 in the brain, kidney and liver, a reduced exposure to high FiO2 in the infants’ lungs tissues by itself may reduce oxidative stress independent of organ oxygenation.
The magnitude of the effect and/or the differential effects of various closed-loop FiO2 control systems on time below, within, or above the SpO2 target range was reported to be very different.18 This may be related to differences in the patient populations studied, in the SpO2 target chosen by the investigators, in differences in the algorithm, in the ventilator hardware and to performance characteristics of the staff during control condition.
Our study is limited by the small sample size and by the design of the study, which precludes assessment of long-term clinical effects. As closed loop FiO2 control becomes available for clinical use, it is extremely important to further elaborate the long-term outcomes in a larger sample size.18 The reduction of workload for nursing and respiratory staff is an additional benefit, which we believe is extremely beneficial for the babies’ overall care. Furthermore, by setting appropriate alarm limits for FiO2 to detect eventual deteriorations of the infant’s respiratory status, it should be possible to avoid that the automated FiO2 system is masking such deterioration. A thorough introduction and comprehensive knowledge of an automated FiO2 control system is clearly indicated for safe routine use in premature infants.
Closed-loop inspired oxygen concentration using the SPO2C mode of the Sophie infant ventilator significantly increased time of arterial SpO2 within the intended range in VLBWI and decreased the need for manual adjustments when compared with the routine adjustment by staff members over a period of 24 hours. No significant difference in time within a prespecified regional cerebral, hepatic or renal tissue oxygen saturation range was detected, which may be related to the postnatal age of the infants studied. Automated FiO2 control may have the potential to decrease morbidity in high-risk premature infants, but long-term effects and safety of this closed loop system of FiO2 control should be evaluated in large randomised controlled trials in VLBWI.
The authors would like to thank the German Federal Ministry of Economic Affairs and Energy (BMWi).
Contributors MG (first author) collected data, provided and cared for study patients. MW, MRM and WB (coauthors) served as scientific advisors, reviewed the study proposal. HH served as scientific advisor.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent Parental/guardian consent obtained.
Provenance and peer review Not commissioned; externally peer reviewed.