INTRODUCTION

TOP ABSTRACT INTRODUCTION CASE REPORT DISCUSSION REFERENCES

Almost half a century ago, Cross and Oppé (1) expressed the view that periodic breathing in the infant "may be induced by hypoxia and relieved by high oxygen. . . ." The conclusion that periodic breathing can be abolished by oxygen has been amply confirmed in infants (2). Likewise, Cheyne-Stokes or periodic breathing in adult humans is reduced or suppressed by oxygen, a fact first reported at the turn of the century (5), and confirmed in the majority of studies that have examined the question since that time (6). As a result of the consistent effect of oxygen, it is now being recommended as a treatment for Cheyne-Stokes breathing patterns in adults (12).

We report on an infant admitted to hospital with recurrent cyanosis in which oxygen exacerbated the instability of periodic breathing. Return of this infant to air resulted in a more profound desaturation than those that led to the use of oxygen in the first place. The destabilizing effects of oxygen in this infant are reminiscent of breathing patterns induced by oxygen in the lamb (13), effects for which we have provided an explanation that highlights a potential danger in the use of oxygen to treat severe desaturation during periodic breathing in the infant.

	CASE REPORT

TOP ABSTRACT INTRODUCTION CASE REPORT DISCUSSION REFERENCES

The case baby was a twin born at 34 wk gestation with a normal Apgar score and a normal birth weight of 1.51 kg. Her early neonatal course was complicated by recurrent central apnea and periodic breathing and confirmed gastroesophageal reflux on pH monitoring (pH < 4 for 10.5% of a 24-h period, against the normal value for our clinic of 3.4%). She was discharged home on postnatal Day 59 on infantile gaviscon (a food thickening agent for treating gastro-esophageal reflux) and an apnea mattress (capacitance movement sensor) but was readmitted on three occasions with recurrent apnea requiring stimulation. On the second occasion a herpes simplex type 1 virus was isolated from a nasopharyngeal aspirate, but she was negative to respiratory syncytial virus and pertussis.

At the time of her third admission on Day 104, she was tachypneic with episodes of coughing and subsequently an adenovirus type 1 was isolated. She continued to have prolonged central apneas longer than 20 s, with bradycardias and central cyanosis. On Day 105 (when she had a normal weight of 4.7 kg) and on Day 107, central apneas were confirmed in a daytime sleep study. These apneas were mostly of a periodic nature, and occurred without evidence of acid regurgitation. Her events became less frequent as her coughing improved, until on Day 113 a repeat afternoon sleep study showed no periodic breathing and no apnea lasting more than 10 s. She was discharged home on Day 119 on a cardiorespiratory monitor and oral theophylline. No further hospital admission was required. Apneic events continued, generally associated with a respiratory infection with coughing, but these became less frequent and less severe. At 8 mo of age, she was diagnosed with eczema and asthma. Monitoring ceased at 2¹/₂ yr of age. Review at 2 yr and 9 mo confirmed a normal development.

During the sleep study on Day 105, ventilation was measured using inductance plethysmography (Respitrace Systems, Ardsley, NY), oxygen saturation was measured with a pulse oximeter (model N-200; Nellcor, Hayward, CA), heart rate (HR) was derived from the electrocardiogram (ECG) signal (Lead II) and the electroencephalogram (EEG) was measured between leads C4-A1. In line with the evidence that oxygen abolishes periodic breathing, and therefore could be expected to prevent the cycles of hypoxemia for which the infant was admitted to hospital, the infant was provided with a flow (4 L/ min) of 100% O₂ delivered via a loose mask held approximately 2 cm away from the face. The first response to inspiration of hyperoxic gas was an extension of the length of each cycle of periodic breathing, with the period of ventilation () being shortened (from 7.7 ± 1.1 s to 5.4 ± 0.6 s; mean ± SD) and the duration of apnea (A) being substantially increased (from 10.3 ± 2.5 s to 34.8 ± 6.8 s), i.e., there was a decrease in the /A ratio from 0.77 ± 0.19 to 0.16 ± 0.04 (Figure 1). After a small number of cycles of periodic breathing, the ventilatory pattern switched to a new cyclic form in which long breathing phases (71 ± 16 s) alternated with long apneas (52 ± 14 s; Figure 1). This new pattern of breathing is similar to what we refer to as episodic breathing in the lamb (see the following section). The switch from air to hyperoxic inspired gas was performed on three occasions when the infant's breathing pattern was clearly periodic, and in all three cases episodic breathing ensued.

View larger version (26K): [in this window] [in a new window]   Figure 1.   Recording of unstable breathing during quiet sleep in the infant at 105 d postnatal age and the effect of supplementary oxygen, showing EEG, HR, respiratory movements of the chest and abdomen (Resp. Chest and Resp. Abd), and oxygen saturation (SpO2). In panel A, three cycles of periodic breathing occurred before supplementary oxygen was provided at the infant's nostrils. The fourth and fifth cycles were extended before the episodic pattern appeared. Panel B shows the lengthy breathing and apnea phases that characterize episodic breathing. Note that when the inspired gas reverted to air in panel C, there was a profound desaturation during the first apnea followed by a return to periodic breathing with less pronounced desaturations.

Soon after institution of hyperoxic therapy, arterial oxygen saturation remained at 100%, even during the long apneas of episodic breathing (Figure 1). Hyperoxia therefore solved one of the problems for which it was employed, even though gross respiratory instability persisted. At the time, we could not explain the genesis of the episodic pattern, nor could we account for the very pronounced arterial desaturation that occurred on all three occasions when the infant was returned to air during episodic breathing (Figure 1). On the day after the study, illustrated in Figure 1, a further sleep study was performed. Similar unstable breathing patterns were observed, and again these were accompanied by pronounced arterial desaturation.

Lamb Study

An unstable breathing pattern, involving cycles of breathing and apnea which last approximately 10 s, can readily be produced after a period of hyperventilation in lambs. We have shown that this pattern is dependent upon peripheral chemoreceptor activity generated by cyclic desaturation of the arterial blood (14). This pattern of breathing occurs spontaneously in the human infant at sea level (3, 15, 16), and in the adult at altitude (17), and is usually referred to as periodic breathing.

In the lamb during periodic breathing, switching the inspired gas from air to hyperoxic gas caused an extension of the duration of the periodic breathing cycle and caused a fall in the /A ratio (Figure 2). In a group of 10 lambs (13), cycle time increased from 9.4 ± 0.4 s to 16.4 ± 1.8 s, whereas /A ratio decreased from 1.5 ± 0.3 to 0.3 ± 0.1. This effect represents a destabilization of the periodic breathing pattern (6) and is identical to that shown in Figure 1 for the human infant. As we explain later, this effect is a consequence of inspired oxygen increasing both the gain of the peripheral chemoreceptor feedback loop and the oxygen store in the lung. In most lambs, after a small number of cycles of periodic breathing, the respiratory pattern became continuous. However, in a small proportion of our lambs, treatment with oxygen resulted in a radical transformation of the respiratory pattern after 2 or 3 cycles of periodic breathing. The new pattern was characterized by breathing phases of approximately 60 to 90 s duration alternating with repetitive apneas lasting up to 60 s (Figure 2; 13). This pattern resembles that produced by oxygen in the human infant (Figure 1).

View larger version (20K): [in this window] [in a new window]   Figure 2.   Unstable breathing patterns in the lamb, showing carotid arterial blood pressure, HR, ventilation, and SpO2. In panel A, 4 cycles of periodic breathing are shown before oxygen replaced air as the inspired gas. Oxygen caused an extension of the periodic breathing cycle ( panel A) before the development of episodic breathing in which a period of breathing lasting approximately 60 to 90 s alternated with an apnea of approximately 60 s duration ( panels A and B). The numbers at the top of the ventilatory signal in panel B correspond to the PCO2 level measured in a sample of jugular venous blood taken at the start and end of the two ventilatory phases shown.

	DISCUSSION

TOP ABSTRACT INTRODUCTION CASE REPORT DISCUSSION REFERENCES

On the basis of our lamb study (13) we can explain the action of oxygen in lengthening periodic breathing cycles and in generating episodic breathing in the infant. Further, we can suggest an explanation for the severe hypoxemia that developed when the infant was returned to air.

Lengthening of the Periodic Breathing Cycle

During periodic breathing in the human infant and lamb, when oxygen replaced air as the inspired gas, its first effect was to lengthen the periodic breathing cycle and to reduce the / A ratio. These effects can be understood by recognizing that the breathing phase of each cycle is initiated when Pa_O2 is falling, and peripheral chemoreceptors are being increasingly stimulated. Once the threshold for breathing is reached, breathing leads to a rise in alveolar PO₂ (PA_O2) and Pa_O2. After a circulation delay, PO₂ at the carotid bodies rises, peripheral chemoreceptor drive to breathing falls, and apnea develops when this drive falls below the apneic threshold. Once breathing stops, Pa_O2 begins to fall and the cycle repeats.

Hyperoxic inspired gas affects periodic breathing because the rate at which PA_O2 (and Pa_O2, because we can assume that alveolar and arterial oxygen concentrations are in approximate equilibrium) rises after the onset of ventilation is directly related not only to the level of ventilation but also to the difference between the existing PA_O2 and the PO₂ of the inspired gas (PI_O2). The greater the PI_O2, the faster the rise in PA_O2 for a given level of ventilation. The result of inspiring hyperoxic gas is that peripheral chemoreceptor drive becomes subthreshold for breathing more quickly than occurs when air is breathed, leading to a shortening of the breathing phase of the cycle. Hyperoxic inspired gas also raises PA_O2 and Pa_O2 to a higher level than when air is inspired so that when apnea starts, Pa_O2 declines from a higher level and takes longer to fall to the threshold at which breathing is initiated. Thus, the duration of the apneic phase of periodic breathing increases. The final effect of hyperoxic inspired air occurs via the decline in /A ratio, which leads to a significant hypoventilation, thereby causing tissue CO ₂ levels to rise more quickly than during periodic breathing in air. Accordingly, respiratory drive generated at the central chemoreceptors rises more quickly when periodic breathing occurs in oxygen than in air, resulting in an earlier restoration of continuous breathing. Hyperoxia therefore promotes stable breathing, but it is important to recognize that before it does so it transiently destabilizes breathing (13).

The initial destabilization of periodic breathing in hyperoxia, resulting from its action in increasing the gain in the chemoreceptor control loop (13), is not peculiar to the lamb. An analogous effect is seen in adult cats when the amplitude of the phrenic nerve discharge is used to set the tidal volume delivered by a mechanical ventilator. In this experimental arrangement, the loop gain of the respiratory controller may be increased by increasing the tidal volume delivered for a given level of phrenic discharge. When loop gain is increased in this manner it causes unstable breathing, and further increases in loop gain cause a progressive decrease in /A ratio (6). The experimentally observed destabilizing effect of increased controller gain in lambs and cats is consistent with the predictions of a mathematical model of respiratory control which incorporates the difference between P I_O2 and Pa_O2 as a multiplying factor in the loop gain equations used to predict respiratory behavior (20).

Development of Episodic Breathing

In the case infant, and in a small proportion of our lambs (13), episodic breathing developed soon after PI_O2 was raised during periodic breathing, suggesting that the mechanism underlying this pattern is similar in the two species and that the peripheral chemoreceptors are not involved. Apparently, therefore, the source of respiratory drive during episodic breathing resides in the central respiratory controller.

We recently proposed a model of the central respiratory controller that explains the key features of episodic breathing in the lamb (13). The model incorporates two characteristics that predispose to unstable breathing. The first, which is observed when breathing starts after a period of apnea, manifests as an immediate jump in ventilation from zero to a level that exceeds that required for eucapnia, resulting in a decline in PCO₂. This behavior, which we call jump, is equivalent to an infinite gain segment in the respiratory controller, a nonlinear feature which can produce unstable behavior in control systems. In the ventilatory system, such a feature can be expected to give rise to abrupt switches between hyperventilation and apnea. However, an infinite gain segment cannot explain the long duration of the breathing and apneic phases of episodic breathing. To do so, our model incorporates a second feature that we refer to as hysteresis in the threshold that governs when apnea begins (apneic threshold) or is terminated (ventilatory threshold). The experimental evidence supporting hysteresis is shown by the jugular venous P CO₂ values shown above the ventilatory signal in Figure 2. These values represent PCO₂ at the central chemoreceptors (13) and they show that the apneic and ventilatory thresholds, at least under the conditions of hyperoxemia, differ by 7 to 8 mm Hg in lambs in which episodic breathing occurred.

Our graphical model of the mechanism underlying episodic breathing is shown in Figure 3. The model assumes that central chemoreceptor PCO₂ rises and falls linearly between the apneic and ventilatory thresholds. At the ventilatory threshold, ventilation jumps abruptly from zero to a peak level from which it falls gradually in a manner consistent with a declining central chemoreceptor PCO₂. Once the apneic threshold is reached, apnea begins and persists until central chemoreceptor PCO₂ rises to the ventilatory threshold. The hysteresis in the threshold for apnea and ventilation has the important consequence that apnea continues until metabolic production of CO₂ raises central chemoreceptor PCO₂ to the ventilatory threshold. Thus, apnea duration is determined by the difference in the apneic and ventilatory thresholds, and by the rate of metabolic production of CO₂. Similarly, once ventilation begins it continues until central chemoreceptor PCO₂ falls to the apneic threshold. Thus, the length of the ventilatory phase of episodic breathing is determined by the extent to which ventilation is hyperpneic (exceeds that required to remove metabolically produced CO₂).

View larger version (16K): [in this window] [in a new window]   Figure 3.   Proposed model of the central respiratory controller, showing the genesis of episodic breathing, an unusual output of the controller that results when there is a large difference in the central PCO2 level at which breathing stops (the apneic threshold) and the central PCO2 level at which breathing restarts after a period of apnea (the ventilatory threshold). Central PCO2 at each instant in time is indicated with a small letter and the ventilation at that instant with the same letter capitalized. Initially at a, central PCO2 is below the ventilatory threshold (Pv,th) and breathing is suppressed (at A). During the apneic period, PCO2 rises centrally until it reaches Pv,th at b. Breathing then starts under the control of the central rhythm generator and ventilation ( I) jumps abruptly from B to B'. I now decreases progressively from B' to C as central PCO2 falls until the apneic threshold is reached at c when breathing turns off. During the apneic interval, PCO2 rises until breathing turns on again at d. The cycle then repeats, generating the pattern characteristic of episodic breathing (from Reference 13; reproduced with permission of the Journal of Physiology).

A number of studies have reported an association between apnea and respiratory infection in babies, especially in babies delivered preterm (21). Although the cause of apnea in these babies has not been established, it was recently shown to be significantly associated with an elevated blood PCO₂ (22). Interestingly, we have found that episodic breathing occurs only in lambs in which the ventilatory threshold lies at high PCO₂ levels (13), suggesting that the episodic pattern is expressed only if there is a preexisting central depression of ventilation. This raises the possibility that the viral illness associated with episodic breathing in our case baby may have caused a reduction in ventilatory drive.

A final observation in the infant that needs to be explained is the profound desaturation on return to air during episodic breathing (Figure 1). We have no data in the lamb that would provide an experimentally based explanation for this observation, but the depth of the desaturation demands that we attempt an explanation. We propose that the return to room air exposes the infant to a dangerous feature of a respiratory controller with hysteresis in its mode of operation. When we compare the first ventilatory phase of episodic breathing after return to air to the preceding one in hyperoxia, it is clear that they have a similar pattern and duration, suggesting that the apneic threshold was unaffected by return to air. It is also to be expected that the ventilatory threshold for CO₂ would remain 7 to 8 mm Hg above the apneic threshold, just as was the case during hyperoxia. At the start of apnea, then, ventilatory drive from the central chemoreceptors is substantially subthreshold and will not reach threshold for up to 60 s. During this time, a desaturation of the arterial blood must occur, because PA_O2 at the start of apnea can be expected to approximate 100 mm Hg in air compared with several hundred mm Hg in hyperoxia. As apnea continues, and desaturation becomes more severe, drive from the peripheral chemoreceptors ultimately becomes sufficient to initiate breathing. What the observations in the infant suggest is that in the absence of central chemoreceptor drive the required peripheral chemoreceptor drive can be generated only by an intense desaturation.

In conclusion, we have provided an explanation for the lengthening of the periodic breathing cycle when hyperoxic inspired gas replaced air in the case study. Our explanation reveals that hyperoxia may transiently exacerbate an unstable breathing pattern, a finding we have discussed in full recently (13), but one that is not widely recognized. We have also put forward a model that accounts for the genesis of episodic breathing in this infant in hyperoxia. Ours is the first report of this form of breathing in the infant, although we have already commented on the similarity between episodic breathing and the respiratory pattern exhibited in Rett syndrome (13). Finally, we have sought to explain how a return to air during episodic breathing in the infant can lead to a pronounced desaturation of the arterial blood. On the basis of our findings and conclusions, we suggest that oxygen therapy for cyclic hypoxemia should be handled with care, and with knowledge of the potential dangers.