CROSSOVER FROM AUTOMATED TO MANUAL TITRATION OF FiO₂ IN THE NICU: IS THERE A TRANSITION LAG?

Overview

Effective management of oxygenation of preterm infants in critical care profoundly impacts their outcome. Nurses are challenged to titrate the inspired oxygen in response to constant cardiopulmonary instability. Closed loop control of inspired oxygen based on continuous monitoring of oxygen saturation is just becoming available. Evaluating the relative effectiveness of closed loop control systems is complicated by the wide variability in manual control by nurses. This analysis explored the possibility of a lag in effective control associated with the transition from closed loop to manual control using data from a clinical crossover trial. A short but marked lag phase was detected. It is, however, unlikely to have impact on clinical care or crossover studies. Its presence highlights the anticipative nature of closed loop control as contrasted to the observative nature of manual control.

Keywords:
Neonatal critical care, oxygen control, closed loop oxygen control

Introduction

After decades of promising research the closed loop control of inspired oxygen (FiO₂) based on pulse oximetry (SpO₂) is finally becoming available in the neonatal ICU. [1, 2, 3] Its routine clinical adoption will dramatically reduce the challenging and frustrating task nurses face in adjusting inspired oxygen in preterm infants during their weeks of respiratory support. [1, 2, 3] Numerous studies of preterm infants have shown the criticality of maintaining oxygen saturation in a “sweet spot” to balance the risk of bad outcomes associated with both hyperoxemia and hypoxemia. [4] Therefore better control of SpO₂, that has become associated with closed loop control, further offers the promise of improved outcomes. [1, 3]

The uptake of oxygen from the alveoli to the pulmonary capillaries is quite fast. Further the time constant of washing in inspired gases throughout the lung is certainly less than a minute. [5] In the neonatal ICU numerous factors have impact. The equipment dead space and time constants associated with changes in SpO₂ in response to an adjustment of FiO₂ in the ventilator were recently characterized during respiratory support in the neonatal ICU. [6] Of course the lag phase, from adjustment to first response, varies primarily according to the ventilator, breathing circuit dead space and other ventilator system considerations. Further, as would be expected, there is also variation in the time constant and gain associated with a change in inspired and arterial oxygen, based on the respiratory mechanics and pathophysiology of the infant. Most automated FiO₂-SpO₂ control systems measure SpO₂ and adjust FiO₂ much faster than the infant-ventilation response. [3] The opposite is true during manual control. However, one small study of manual and automated control suggested that quicker response, whether associated with auto or dedicated manual control, resulted in more effective control. [7] Nevertheless quicker increases must also be associated with quicker decreases when faced with hyperoxemia.

Some automated systems have also been characterized as having an anticipative component in their control algorithms. [3] As an example, they might consider the rate of change of the SpO₂ in making a decision about the next FiO₂ adjustment. In contrast, routine manual control would be characterized as observative. That is, the nurse waits to observe the complete response to a change in FiO₂ before considering the need for another change.

Physiological crossover studies should always consider the kinetics of the therapeutic intervention, and employ an adequate washout period between interventions. Nearly all of the evaluations of automated FiO₂-SpO₂ control utilized a physiological crossover design. However, because the SpO₂ response to changes in inspired oxygen was known to be quite rapid compared to the duration of the intervention periods, washout periods in these studies have been considered unnecessary.

The aim of this study was to determine if there was a clinically relevant crossover effect when switching from automated to manual FiO₂-SpO₂ control. We speculated that the transition from automated to manual control, by virtue of its observative character, would result in a distinct transition.

Methods

Source data from a large crossover trial was used. [8] This trial enrolled 80 preterm infants at 9 centers. The order of intervention was randomized, with 40 receiving manual control after automated control. The database included the subjects’ demographics as well as SpO₂ data sampled every 5 seconds for the two 24-hour intervention periods. Five of the 40 cases that transitioned from automated control to manual control were randomly selected. In addition, for the purpose of subjective comparison, 5 of the other 40 cases that transitioned from manual to automated control were also randomly selected. SpO₂ data were extracted for the hour before and after the transitions between automated and manual control.

The system used in the van Kaam trial [8] was the AVEA-CLiO2 (CareFusion, Yorba Linda CA, USA). This system had been commercially available in Europe for a number of years. The control algorithm continuously monitors FiO₂ and SpO₂ and makes a decision every second about whether a change in FiO₂ is needed. The decision is based on a set of rules that vary depending on whether the SpO₂ is below, within or above the set SpO₂ control range. The rules, while varying depending on that range, all consider 6 parameters. These are the rate of change of SpO₂, the direction of change, the time outside the target range, the absolute deviation from the target range, the reliability of the pulse oximetry signal and the baseline FiO₂. More details of the control algorithm and system performance are available. [3, 8]

Mean SpO₂ was prospectively defined as the primary metric of control. The proportion of times with possible hypoxemia and possible hyperoxemia were selected as secondary endpoints, that might be associated with a shift in the mean. Possible hypoxemia and hyperoxemia were define as SpO₂<86% and SpO₂>96% when not inspiring room air, respectively. The presence of a relevant transition/washout was subjective, determined graphically. A 2-minute rolling mean of the SpO₂ was used to filter out the normal rapid variation in SpO₂ to facilitate identification of a transition effect. A two-minute averaging was selected as nearly all desaturation episodes are less than 1 minute. [9]

Differences in the mean SpO₂ were evaluated using a pooled two-sample t test. Differences in hyperoxemia and hypoxemia were evaluated using a Z-test of proportions. A P value of <0.05 was considered statistically significant. Statistical analyses were conducted using XLStat (v16.6.03, Addinsoft USA).

Results

The 10 subjects selected were preterm infants studied in 4 neonatal ICUs. The interquartile range of their demographics were as follows: estimated gestational age (25‒26 weeks), age and weight when studied (16.5‒27.8 days) and (0.97‒1.22 kg). Most were receiving noninvasive respiratory support (6 continuous positive airway pressure, 2 intermittent positive pressure). During this two-hour analysis period the interquartile range of the SpO₂ was 91‒96% in the automated to manual group and 90‒95% in the manual to automated group.

Figures 1ab present the rolling average of the SpO₂ over the two-hour analysis periods. As would be expected even with the rolling average the SpO₂ is not stable. The transition to manual from automated control is clear with a marked drop in SpO₂ (Figure 1a). A transition period of 10 minutes was selected after creating the charts. The slope of the rolling average at the boundaries to the 10-minute transition period suggests a precipitous transition. In contrast (Figure 1b) no transition period is apparent when switching to automated from manual control.

The actual means for all the 5-second data points in two 10-minute transition periods are compared to the subsequent 50 minutes and shown in Table 1. As suggested in Figure 1a, during manual control the mean SpO₂ in the transition is significantly lower (2.4% SpO₂, P<0.001). In contrast, consistent with Figure 1b, there was no apparent difference in mean SpO₂ transitioning to automated control. The variation of SpO₂ (standard deviation) was similar among the periods.

The lower mean SpO₂ during the manual transition also impacted the distribution at SpO₂ extremes. The percent time with a SpO₂ reflective of a risk of hypoxemia was higher during the transition (16.9% vs 7.2%, P<0.001) and correspondingly less with a risk of hyperoxemia (15.7% vs 25.9%, P<0.001). The percent time with SpO₂ between these two extremes, normoxemia, was however unchanged (67.4% vs 66.9%, P=ns).

Discussion

The maintenance of SpO₂ at the beginning of manual control of FiO₂ was different than during ongoing control, as hypothesized. A similar transition was not seen when starting automated control. In the clinical environment drops in SpO₂ are usually a result of disordered breathing. Thus maintaining desired SpO₂ levels necessitates intermittent increases in FiO₂. During the manual control transition period the SpO₂ was lower, consistent with the observational nature of manual control.

We found the transition period was about 10 minutes. Fathabadi et al, reported on the neonatal SpO₂ response to changes based on 580 isolated increases in FiO₂ in 47 subjects. [6] They reported a typical delay of 22 seconds (IQR 8‒40) before SpO₂ started to increase. They also found that the 95% response to an increase in FiO₂ was 39 seconds (IQR 7‒105). While they reported considerable variation among episodes and subjects, this data suggests that two or three adjustments, each occurring after observing the response to the last adjustment, nominally a minute, would be much shorter than the 10 minutes that was identified as the transition period. We speculate therefore that during manual control the periods of observation from initiation and subsequent FiO₂ adjustments are much longer than the physiological responses. This is consistent with the multitasking required of neonatal nurses.  

While the SpO₂ transition to manual adjustment was marked, the rolling average SpO₂ stayed within the range associated with normoxemia (SpO₂ 86‒96%). The shift to lower levels resulted, nevertheless, in significant differences in SpO₂ exposure. Specifically, there was an increased exposure to potential hypoxemia and decreased exposure to potential hyperoxemia. Differences of this magnitude are certainly clinically relevant, if they were to persist for a significant amount of time. [9, 10] It is unlikely, however, that infrequent transitions associated with shift or staffing changes would occur frequently enough to impact accumulative exposure. However some part of the differences reported with staffing levels might be associated with transitions associated with attending to particularly unstable infants. [11]

It appears that the transition period would not have a material effect on the results of crossover studies of FiO₂-SpO₂ control systems. As an example, even in a study with 4-hour intervention periods a 10-minute transition would represent only 4.2% of the intervention time. Studies more typically have utilized 12 or 24-hour intervention periods.

This study has some limitations. The analysis is based on the response of 5 subjects. While they were picked at random from a larger population, selection of different subjects might have resulted in different results. The analysis also used the pooled response of the subjects, a larger study evaluating paired within-subject differences would provide information about the variation among subjects. Nevertheless the differences we found were clinically relevant and highly statistically significant. This data came from a controlled trial in which the caregivers knew their manual control was being studied. In such circum-stances one would expect the control would better, and thus the effect less pronounced than during routine care. The actual transition effect associated with shift changes should be studied at centers with appropriate patient data monitoring systems.

Conclusion

There appears to be a lag phase with a drop in SpO₂ at the beginning of manual titration of FiO₂. This transition phase is longer than just the physiological response would predict. Its duration and magnitude are not sufficient to have any impact on patient risk or study design.

Acknowledgement

The work was not supported by a research grant or contract.

Thomas Bachman MSc

Faculty of Biomedical Engineering

Czech Technical University in Prague

Faculty of Biomedical Engineering

Department of Biomedical Technology

nám. Sítná 3105, CZ-272 01 Kladno

&

Economedtrx

Lake Arrowhead, CA USA 92352-1269

E-mail: TBachman@me.com

Phone: +1 909 337-0828