INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The hyperoxic test (HT) examines peripheral chemoreceptor function by inducing "physiologic chemodenervation." The drop in ventilation after a change in the inspired fraction of O₂ from normoxia to hyperoxia is believed to reflect an acute reduction in peripheral chemoreceptor input and, therefore, the strength of the peripheral chemoreceptor drive (1).

Various durations of O₂ inhalation have been used for HTs in infants, ranging from one or two breaths (3, 4) to 30 s (5) or longer (9). When O₂ inhalation was restricted to one or two breaths, the initial drop in ventilation after the hyperoxic stimulus was evaluated as the difference in ventilation variables during hyperoxia as compared with normoxia, expressed as the percentages of values during normoxia (3, 4). However, uncertainty exists about whether peripheral chemoreceptor input is completely eliminated by a hyperoxic stimulus restricted to one or two breaths (12). When the duration of 100% O₂ inhalation was 30 s, the ventilatory response to hyperoxia was calculated as the mean ventilation reduction during the hyperoxic period (5). However, it has been shown that the initial drop in ventilation during hyperoxia is followed by a rise, which reaches values above the normoxic level after 1 to 2 min (2, 9). Thus the mean change in ventilation during 30 s of hyperoxia reflects both the initial drop caused by "peripheral chemodenervation" and the early phase of the subsequent increase in ventilation related to the metabolic response to hyperoxia in infants (2). Therefore, the mean change in ventilation over a 30-s hyperoxic period would be expected to underestimate the strength of peripheral chemoreceptor function. Furthermore, it is reasonable to assume that the ventilatory response to HT as assessed by this method may be affected by the intrinsic variability of both the peripheral chemoreceptor drive and the metabolic response. A single study has evaluated the reproducibility of the ventilatory response to hyperoxia (6). In this study, the ventilatory response to a 30-s hyperoxic period, calculated as the mean change in ventilation, was reproducible in only half of the study infants (6).

The present study was undertaken to reinvestigate the method of calculation of the ventilatory response to 30-s HT with regard to the reduction in peripheral chemoreceptor drive, and to assess the reproducibility of the ventilatory response to HT. Breath-by-breath analysis allows the determination of the response time (RT) of the ventilatory response to changes in chemical stimuli (7, 8, 13). In the HT test, the RT is defined as the time from hyperoxia onset to the first significant ventilation decrease during hyperoxia as compared with normoxia (8). Evaluation of the ventilatory response at the RT during O₂ inhalation for 30 s should ensure that the values obtained reflect complete elimination of peripheral chemoreceptor activity, an important advantage over hyperoxia restricted to one or two breaths. Furthermore, since the ventilatory response at the RT results only from "peripheral chemodenervation," we hypothesized that it would be greater and more reproducible than the mean percentage change during the 30-s hyperoxic period. A reproducible test for assessing peripheral chemoreceptor function may prove clinically useful since peripheral chemoreceptor dysfunction has been implicated among the pathogenic mechanisms of sudden infant death syndrome (14).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Eighteen infants were studied. Our study infants were not healthy full-term control infants. Their biometrical characteristics and neonatal history are summarized in Table 1. Six were born prematurely. Three who had respiratory distress syndrome at birth were not mechanically ventilated, but received oxygen therapy during the first three postnatal days. All 18 infants were hospitalized in the pediatric department of our university hospital, from which they were due to be discharged within 1 to 3 d. All were free from neurologic, cardiac, or respiratory symptoms at the time of testing, and none had received medications known to affect cardiorespiratory variables or sleep organization. The study protocol was approved by our institutional review board. Parents gave their informed consent for participation of their infants in the study.

Protocol

The infants were studied in the supine position during a morning nap, without sedation. Room temperature was 22 to 24° C. Tests were done 1 h after feeding, during quiet sleep as assessed based on behavioral observations (15). A face mask attached to a pneumotachograph was gently placed over the face of the infant. Care was taken to avoid leakage. At least 1 min was allowed to elapse before data storage initiation to ensure that no changes occurred in the state of alertness. Two HTs were performed. Each HT included a 30-s normoxic control during which infants breathed room air followed by a 30-s hyperoxic period during which infants inspired 100% of O₂ from a bag and expired to room air. The interval between the two onsets of hyperoxia was 2 min. Therefore, the duration of the normoxic period preceding HT2 was 90 s. Tidal volume was displayed breath by breath on the computer screen during the test.

Equipment

The experimental setup allowed measurement of respiratory flow, determination of O₂ and CO₂ fractional concentrations near the infant's mouth, and use of a respiratory valve to produce a 100% fractional concentration of inspired O₂ during the HTs (16). Respiratory flow was monitored using a screen pneumotachograph (Statice Santé, France) attached to a face mask (silicon mask #0; GlaxoWellcome, France) fitted to the infant (17). The stainless steel screen was placed between the two portions of the pneumotachograph. Two linearizing screens were placed 12 mm from the resistive screen to ensure a linear response to a rate of 500 ml/s (18). The differential-sensing ports were located 5 mm from the resistive screen. The apparatus had a flow resistance of 0.3 kPa/L/s and a total instrumental dead space of 4 ml. The pneumotachograph was calibrated during normoxia and hyperoxia. The flow signal was converted to an analog signal via a pressure transducer (Validyne DP-45-16, ± 2 cm H₂O; Validyne Corp., Northridge, CA) driven by an amplifier (Carrier 13-G4615; Gould Inc., Cleveland, OH). A two-way electric valve secured to the pneumotachograph was controlled by a software program to produce the 30-s square wave 100% fractional concentration of inspired O₂. The response time of the respiratory valve was 30 ms. The valve controlled two pathways, one for ambient air and the other for 100% O₂. The oxygen was in a 10-liter Douglas bag filled from a cylinder (Air Liquid Corp., France). At the beginning of each 30-s of hyperoxic period, the investigator activated computer detection of the end of the next expiration, which was followed by inspiration of 100% O₂ from the bag. The valve was switched back to the other pathway at the end of each inspiration so that expiration occurred to room air. Inspired and end-tidal gas levels were determined using a fixed magnetic field mass spectrometer (MGA-1100; Perkin-Elmer, Pomona, CA) capable of simultaneously detecting O₂ and CO₂. The gas mixture was sampled at a rate of 18 ml/min through a 1-m-long capillary inserted into the lumen of the pneumotachograph so that its tip was near the infant's mouth. The mass spectrometer had a 300-ms transport time lag and a 100-ms 95% response time. Analyzer analog outputs for O₂ and CO₂ were connected to signal conditioners (DC-57-1340; Gould Inc.). Respiratory flow and fractional concentrations were digitized at a sampling rate of 200 Hz per channel by an acquisition board (DAS1600; Keithley-MetraByte, Taunton, MA). The electrocardiogram was recorded using self-adhesive electrodes placed on both sides of the chest anteriorly and on the abdomen. Heart rate (HR) was calculated from R-R intervals measured on the electrocardiogram after butterworth filtering and peak detection (19). Oxygen saturation was monitored throughout the experiment using a pulse oximeter with a neonatal sensor (SensorMedics Corp., Anaheim, CA) in beat-to-beat mode.

Data Analysis

Respiratory flow was integrated for breath-by-breath measurements of the following respiratory variables: tidal volume (VT), inspiratory time (TI), expiratory time (TE), total duration of the respiratory cycle (Ttot), respiratory frequency (f), duty cycle (TI/Ttot), minute ventilation (E), and mean inspiratory flow (VT/TI). Inspired (FI_O2, FI_CO2) and end-tidal (FET_O2, FET_CO2) fractions of O₂ and CO₂ were detected breath by breath, taking into account the lag time of the mass spectrometer. CO₂ production was calculated as follows: CO₂ = E (FET_CO2  FI_CO2). Data obtained during normoxia and hyperoxia were preprocessed using the following criteria to discard respiratory cycles with apneas or sighs: inspiratory and/or expiratory tidal volume outside the mean VT ± 2 SD of all recorded data; and TI (or TE) outside the mean TI ± 2 SD (or mean TE ± 2 SD). After preprocessing, breath-by-breath data were divided into groups of four breaths (8).

Ventilatory response to hyperoxia. For all the measured respiratory variables, two analyses were performed during hyperoxia. (1) The mean changes in respiratory variables during hyperoxia were expressed as the percentages of the corresponding mean values during normoxia (5). Respiratory variables were compared between normoxia and hyperoxia using paired t tests with the significance level set at < 0.05. (2) The RT was determined as the time from hyperoxia onset to the first statistically significant decrease in ventilation, determined based on periods of four breaths (one-way analysis of variance). Post-hoc simple comparisons were done using Student's t test with p values < 0.05 considered significant.

Reproducibility of the ventilatory response to hyperoxia. We estimated the limits of agreement between percent changes in respiratory variables during the two HTs (20). We also computed inter- and intra-individual coefficients of variation (CV) of the changes in respiratory variables during the two HTs.

All data are reported as means ± SD. Statistical analyses were performed using the Statistica Software package (StatSoft, Inc., Tulsa, OK).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Raw data of gas fractions and respiratory flow from Infant 17 (Table 1) during the first hyperoxic test (HT1) are shown in Figure 1. No data preprocessing was necessary for any of the 18 records since no apneas or sighs occurred during the HTs.

View larger version (34K): [in this window] [in a new window]   Figure 1.   Sampled signals of ventilatory flow, fractional CO2 concentration (FCO2), and fractional O2 concentration (FO2) recorded over two successive hyperoxic tests (Infant 17 in Table 1) and displayed without data preprocessing. Left panels show all recorded data, with onset of hyperoxic tests marked by arrows. Right panels show zoomed data starting at the onset of the first hyperoxic test.

Mean Sa_O2 was 97 ± 1% at baseline and remained 100% during hyperoxia. Mean heart rates during normoxia and hyperoxia in HT1 and HT2 were 98 ± 6, 100 ± 7, 104 ± 8, and 103 ± 7 bpm, respectively. Heart rates were not significantly different between the normoxic periods of HT1 versus HT2 or between the normoxic versus the hyperoxic period of each HT. E s during normoxia were not significantly different between HT1 versus HT2 (384 ± 77 ml/min/kg and 408 ± 119 ml/min/kg, respectively).

FET_O2 increased during the hyperoxic periods but did not reach the inspired values within the 30-s hyperoxic period, as shown in Figure 1. FI_O2 in hyperoxic breaths was 88 ± 6% and 87 ± 5% during HT1 and HT2, respectively. These values are not significantly different.

Ventilatory Response to Hyperoxia

An example of the breath-by-breath changes in respiratory variables (E, VT, TI, and VT/TI) recorded during HT1 (Infant 17 in Table 1) is shown in Figure 2.

View larger version (21K): [in this window] [in a new window]   Figure 2.   Breath-by-breath changes in respiratory variables ( E, VT, TI, and VT/TI) during the first hyperoxic test (Infant 17 in Table 1) extracted from the data of Figure 1. Arrows indicate hyperoxic test onset. Breaths before the arrows occurred during the normoxic period. Open circles indicate the time during the four breaths at which the first significant change in respiratory variables was obtained. Response time was 12 s in Infant 17.

Ventilatory response as the mean percentage change during 30 s of hyperoxia. For pooled data (Figure 3), E, VT, TI, and VT/TI showed significant percentage changes during both HT1 (17 ± 8%; 14 ± 6%; 6 ± 5%; 18 ± 9%) and HT2 (13 ± 7%; 13 ± 11%; 4 ± 9%; 16 ± 13%). TE, Ttot and TI/Ttot, and f did not change significantly during the HTs.

View larger version (28K): [in this window] [in a new window]   Figure 3.   Mean (±SD) percentage changes in E, VT, TI, and VT/TI during hyperoxia as compared with normoxia in all 18 study infants. For each respiratory variable, the figure shows the mean percentage change over 30 s of hyperoxia during HT1 (open bar) and HT2 (closed bar) and the mean percentage change at the response time during HT1 (hatched bar) and HT2 (stippled bar). Significant differences were found for E, VT, and VT/TI (*p < 0.01; **p < 0.001) when the two methods of calculating the response to hyperoxia were compared.

During both HTs, all 18 infants responded to the hyperoxic challenge by decreasing E and VT/TI significantly. Changes in VT and TI differed between infants and tests. During HT1, all but one infant had a significant decrease in VT; TI increased significantly in 11 infants and remained unchanged in six. During HT2, VT decreased significantly in all 18 infants, whereas TI increased significantly in 10, remained unchanged in six, and decreased significantly in two.

Reproducibility of the ventilatory response. The upper limits of agreement of the two HTs were 19% for E, 16% for VT, 22% for VT/TI, and 14% for TI. The intraindividual CVs for E, VT, and VT/TI were 40, 51, and 51%, respectively, and the corresponding interindividual CVs were 50, 64, and 64%, respectively.

Ventilatory response at the RT. RESPONSE TIME. The response time (RT) ranged from 6 to 19 s during both HTs, with no significant difference between HT1 (13 ± 4 s) and HT2 (13 ± 3 s). Intraindividual variability of RT was 21 ± 21% on average. The limit of agreement between the RTs of the two HTs was 10 s (confidence interval, 7 to 13 s).

VENTILATORY RESPONSE. For pooled data (Figure 3), E, VT, TI, and VT/TI showed significant percentage changes at the RT during both HT1 (21 ± 6%; 22 ± 14%; 5 ± 10%; 21 ± 8%, respectively) and HT2 (21 ± 7%; 19 ± 10%; 5 ± 11%; 23 ± 14%, respectively). TE, Ttot, TI/Ttot, and f were not significantly modified during the HTs. Individual percentage changes in E at RT during HT1 and HT2 are shown in Figure 4.

View larger version (45K): [in this window] [in a new window]   Figure 4.   Individual values of percentage change in E at response time calculated during HT1 (open bars) and HT2 (stippled bars) in the 18 study infants, numbered from 1 to 18.

No significant correlations were found between the E decrease and the RT during either HT. All 18 infants showed decreases in E, VT, and VT/TI during both HTs. Changes in TI differed across infants and across HTs. During HT1, TI increased significantly in 11 infants, showed no significant change in four, and decreased significantly in three. Corresponding numbers for HT2 were nine, five, and four.

REPRODUCIBILITY OF THE VENTILATORY RESPONSE. Differences in deviations of E, VT, TI, and VT/TI for each infant were plotted against the corresponding means (Figure 5). The average difference for each variable was not significantly different from zero. The upper limits of agreement between the two HTs were 17% for E, 30% for VT, 27% for VT/TI, and 23% for TI. Intraindividual CVs were 23, 34, and 34% for E, VT, and VT/TI, respectively. Corresponding interindividual CVs were 54, 69, and 80%.

View larger version (14K): [in this window] [in a new window]   Figure 5.   Agreement between the percentage change at the response time during the first and the second HT, for E. Differences between the percentage changes in E are plotted against their means. The solid line indicates the line of identity. The upper and lower dotted lines indicate plus and minus 2 SDs, respectively.

Comparison between the two ventilatory response analyses. Decreases in E, VT, and VT/TI were significantly greater when calculated at the RT than as the mean percentage change during the 30-s hyperoxic period (average differences: 5% for E, 6% for VT, and 5% for VT/TI; p < 0.01). No significant difference was observed for TI. When calculated at the RT, the limit of agreement of the two HTs showed a 2.2% reduction in E when calculated at response time. They were increased on average for the remaining variables (+11% for VT and +7 for VT/TI). At the RT intraindividual CV is diminished on average by 17% for E (p = 0.03), VT (p = 0.03), and VT/TI (p = 0.04). At the RT intraindividual and interindividual CV for TI did not differ between the two calculation modes.

CO₂ fractions and CO₂ production. The mean values and standard deviations of inspired and end-tidal CO₂ fractions under all testing conditions in all 18 infants are shown in Table 2. Mean FI_CO2 and mean FET_CO2 did not change significantly during the 30-s hyperoxic period as compared with normoxia (Table 2). FET_CO2 decreased slightly but significantly at the RT during HT1 and HT2 (p < 0.01) (Figure 6, upper panel). Breath-by-breath analysis of CO₂ (Figure 6, lower panel) showed that CO₂ decreased initially during the 30-s HTs, then increased. When calculated as the mean change, CO₂ decreased significantly versus normoxia, by 12.0 ± 9.0% during HT1 (p < 0.001) and by 12.0 ± 11.0% during HT2 (p < 0.001). CO₂ at the RT was significantly decreased as compared to normoxia, by 24.0 ± 13.0% (p = 0.01) during HT1 and by 20.0 ± 11.0% during HT2 (p < 0.001); these decreases were significantly greater than during the 30-s hyperoxic period (p  0.01).

View larger version (22K): [in this window] [in a new window]   Figure 6.   Breath-by-breath changes in FETCO2 and in CO2 production during the first HT (Infant 17 in Table 1) extracted from the data of Figure 1. Arrows indicate hyperoxic test onset. Breaths before the arrows occurred during the normoxic period. Open circles indicate the time during the four breaths at which the first significant change in FETCO2 and CO2 production was obtained. Response time was 12 s in Infant 17. Dashed lines correspond to mean values of FETCO2 and CO2 production during normoxia.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Our aim was to reinvestigate the method of calculation of the ventilatory response to hyperoxia in infants. We performed breath-by-breath analysis of ventilatory variables and determined the response time (RT) defined as the time from HT onset to the first significant HT-related change in E. We found that the decreases in E, VT, and VT/TI, were significantly greater when calculated at the RT than as the mean percentage change during 30 s of hyperoxia. Furthermore, reproducibility of the ventilatory response to HT was improved as shown by the significant reductions at the RT in the intra-individual coefficients of variation for E, VT, and VT/TI. Peripheral chemoreceptors reset at a higher PO₂ level after birth in newborn mammals (21). This resetting occurs after the first day of life in both preterm (7) and full-term human newborns (6). Because our study infants cannot be considered healthy control infants, our study does not provide normative data on peripheral chemoreceptor function. However, since all our study infants exhibited a significant response to hyperoxia, our study data allowed us to reinvestigate the method of calculation of the ventilatory response. The fall in E was due mainly to a fall in VT, as previously reported in healthy control infants (6). Earlier studies did not evaluate changes in all the components of E during hyperoxia. We found a significant fall in VT/TI, reflecting the decrease in the inspiratory drive. Respiratory rate was unaffected, as previously reported (6). Changes in TI varied across our study infants and tests.

We used a standardized protocol both to perform the HTs and to analyze the respiratory variables. Our unsedated infants were studied during well-established quiet sleep as assessed based on behavioral criteria (15). We applied a face mask attached to the pneumotachograph. Although this type of setup has been shown to increase minute ventilation and tidal volume (22, 23), both effects are transient (24). We consequently started the normoxic test at least 1 min after application of the mask over the infant's face.

We performed breath-by-breath analysis during the HTs. We determined the response time of the ventilatory response to hyperoxia (8) and found that it ranged from 6 to 19 s, with no significant differences between HT1 (13 ± 4 s) and HT2 (13 ± 3 s). However, the response time showed fairly substantial variability between the two HTs in our infants. As pointed out by Ward and Robbins (25), the transient ventilatory response is more difficult to interpret because it involves factors such as dilution and mixing in the lungs and pulmonary circulatory transit times. We evaluated the ventilatory response to hyperoxia at the response time and as the mean percentage change during the 30-s hyperoxic period. Falls in E, VT, and VT/TI were significantly greater when calculated at the response time, consistent with the fact that during hyperoxia ventilation decreases initially, then increases. The mean over the 30-s hyperoxic period is the net result of two different mechanisms, namely, a reduction in peripheral chemoreceptor input and stimulation of ventilation caused by the subsequent increase in metabolism (2).

Breath-by-breath analysis of CO₂ production showed that CO₂ decreased initially. However, CO₂ fell more than E at the response time because of a small but significant decrease in FET_CO2 at the response time. The CO₂ decrease accompanying the E decrease may have been due to the dynamic metabolism-ventilation interactions associated with acute changes in arterial partial pressure of O₂ in infants (26). After the initial decrease, CO₂ subsequently increased during 30-s HTs, consistent with the reported increase in metabolism during more prolonged hyperoxia (2). This combination of an initial decrease and secondary rise in CO₂ resulted in a lower mean CO₂ production during 30-s HTs than during normoxic periods.

Ventilatory response reproducibility has not been extensively studied in human infants. Hertzberg and Lagercrantz (6) calculated the ventilatory response as the mean percentage change over 30 s of hyperoxia as compared with normoxia and found that the fall in E was reproducible in only half a sample of healthy infants of similar postnatal age. We found that the ventilatory response to hyperoxia was reproducible with both calculation methods used in our study. However, calculation of the E decrease at the response time was associated with a marked reduction in the intraindividual coefficient of variation of E, the most useful parameter for clinical investigations. The substantial interindividual variability of E, even when calculated at the response time, may be a limitation. However, comparisons of groups of infants based on HT results may be clinically relevant, as previously reported between preterm infants with bronchopulmonary dysplasia (BPD) and control preterm infants (8). Nevertheless, it may be difficult to perform reliable HT in infants with abnormal lung function; in particular, a false negative ventilatory response may occur when O₂ administration is restricted to one or two breaths because of poor alveolar mixing. The advantage of a 30-s HT with breath-by-breath analysis is that it allows the determination of the response time even in infants with abnormal lung function (8).

Peripheral chemoreceptor drive is of major importance during infancy. Abnormal events during the neonatal period such as hypoxia (21) and nicotine exposure (27) have been shown to delay peripheral chemoreceptor resetting in newborn animals. Delayed resetting of peripheral chemoreceptors has been reported in infants with BPD (8). Infants with BPD (28) and infants born to mothers who smoked during pregnancy (29) have been found to be at high risk for sudden infant death syndrome (SIDS), suggesting that defective chemoreceptor function may contribute to SIDS (14). A reproducible test of peripheral chemoreceptor function may be clinically useful for detecting chemoreceptor dysfunction in infants at high risk for SIDS.