INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Postmortem studies of infants who have died from sudden infant death syndrome (SIDS) have revealed that some of these infants show evidence of chronic hypoxia (1). The brainstems are immature, with increased levels of dendritic spines and synapses (2, 3), and delayed myelination (4), and astrogliosis in areas important for respiratory control (5). In a proportion of these infants the carotid bodies, which are important in ventilatory responses to hypoxia and arousal, have been shown to contain a 10-fold higher concentration of dopamine when compared with infants who died from other causes (8). Intravenous infusion of dopamine in adults (9) and young lambs (10) reduces ventilatory responses to hypoxia. Therefore, if these increases in dopamine content of the carotid body in SIDS victims were also reflected in an increase in dopamine release, then one may expect that these infants would have reduced ventilatory responses to hypoxia. Consequently, the findings that SIDS victims show signs of chronic hypoxia before death, and have abnormalities in their brainstems and carotid bodies, have led to the suggestion that deficits in respiratory control, particularly in peripheral chemoresponses to hypoxia, may be implicated in some SIDS deaths.

It is now well established that maternal smoking, particularly during pregnancy, increases the risk of SIDS (11). Prenatal nicotine exposure in rats has been shown to lead to cell death in the brainstem (19) and postnatal exposure leads to increased dopamine turnover and synthesis in the carotid body (20), suggesting that the peripheral chemoreflex to hypoxia may be affected by nicotine exposure. However, studies investigating the effect of nicotine exposure on respiratory control in animals and humans have been contradictory. Lambs exposed to nicotine postnatally have been shown to have attenuated ventilatory responses to hypoxia but increased responses to hyperoxia (21). These findings conflict with those seen in the rat, where postnatal nicotine exposure reduces the response to hyperoxia (20). Studies investigating prenatal nicotine exposure are also contradictory, with the lamb showing reduced ventilatory responses to hypoxia after nicotine exposure (22) but the rat showing no effect (23). Lewis and Bosque failed to find a difference in respiratory responsiveness to hypoxia between human infants born to mothers who smoked during pregnancy and those who did not (24). However, this study did not control for potential confounding factors related to smoking including social class, maternal age, birthweight, gestational age, and race. These factors may have separate effects on respiratory control and thus may mask any effect of smoking per se. Conversely, a more recent study has reported that infants born to mothers who smoke have blunted ventilatory responses to hypoxia compared with those infants born to nonsmoking women (25). A previous study from our group investigating respiratory responses reported as a subsidiary finding that infants with at least one smoking parent had reduced respiratory responses to changes in inspired oxygen levels (26).

Therefore, evidence suggests that respiratory control deficits may be implicated in SIDS and that nicotine exposure may affect respiratory control. However, results from studies looking at the effect of nicotine exposure on respiratory control in young animals and infants are equivocal. Our aim in this study was to examine the independent effect of maternal smoking on peripheral chemoresponses to changes in inspired oxygen during unsedated night-time sleep in the infant between 8 and 12 wk of age. Confounding factors were controlled by a matched design enabling investigation of the independent effect of smoking. We aimed to test the hypothesis that infants born to mothers who smoke have reduced ventilatory responses to changes in inspired oxygen.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Outline

Healthy smoking and nonsmoking women were recruited in early pregnancy. Urine samples were obtained from the women at 20 wk of gestation for measurement of cotinine, to verify reported smoking habits. Approximately every 5 wk women were contacted by telephone or letter to assess their continued eligibility. After delivery women were visited in the postnatal wards and information was obtained about the delivery and health of the infant. When the eligible infants were between 8 and 12 wk of age they attended the sleep laboratory for overnight measurements of respiratory control and an overnight urine sample was obtained.

Study Design

Power calculations. A previous study conducted within our department found as a subsidiary finding that infants with at least one smoking parent had reduced alternations in inspiratory drive when compared with those of nonsmoking parents (p = 0.09) (26). On the basis of these observations we calculated that 30 infants would be required in each group to detect a difference of 10% in the alternation in inspiratory drive with 91% power.

Matching. A matched study design was chosen to control for confounding factors related to smoking that might also influence respiratory control. In this way confounding factors could be matched out and the independent effect of smoking assessed. Each smoking mother- infant pair was matched on an individual basis to a nonsmoking mother-infant pair for the following factors: social class, maternal age and parity, mode of feeding, birthweight and gestational age of the infant, and infant sex. A weighted scoring system for the matching of the seven factors was employed (Table 1) and a match was accepted if 5 points or fewer were scored. Each smoking mother-infant pair could be matched with more than one nonsmoking mother-infant pair to form a matched set. However, within a matched set all of the nonsmoking mother-infant pairs had to match with one another. Social class was determined and based on the partner's occupation, using Office of Population, Census and Surveys (OPCS) criteria. If the woman was not living with a partner it was based on her occupation. Those who were currently unemployed were classified according to their last job. If they had not been employed in the last 10 yr they were classified as unemployed.

Recruitment

Case notes of women presenting at the antenatal clinic of the Leicester Royal Infirmary were surveyed and those who appeared to be eligible were approached in the clinic at the time of first booking (usually around 13 wk of gestation). The study required that the woman and her partner were both white. Other eligibility criteria were that the woman should (1) be 18 yr of age or over, (2) be fit and well with no significant history, (3) not be taking prescribed medication, (4) have a singleton pregnancy, (5) not be using fertility treatment, and (6) not be using recreational drugs. After the initial approach, if the woman was interested in hearing about the project in more detail she was visited in her home by the research nurse and the formal process of recruitment took place. This included administering a questionnaire seeking information on maternal age, parity, smoking habit, employment status, and social class. If at any stage on or after recruitment the woman's smoking habits became inconsistent, she admitted to using recreational drugs or was prescribed medications (single course of antibiotics excepted) she was excluded from the study.

After delivery, in order for the infant to remain eligible for the study s/he had to be born after 37 wk of gestation, be of normal birthweight, not require any respiratory support, and have no congenital abnormalities. In addition, s/he had to remain well between birth and the time of the study. Minor upper respiratory tract infections were permitted, provided that the infant had recovered at least 2 wk before study.

Written informed consent for the study was obtained from the parent(s). The study was approved by the Leicestershire Health Authority ethics committee. Mothers were not paid to take part in the study but any expenses incurred when their infants were brought to the hospital for testing were reimbursed.

Cotinine Measurements

The urine samples obtained from the mothers at midgestation, and from the infants at the time of testing, were analyzed for cotinine and creatinine. Cotinine was analyzed by an enzyme-linked immunosorbent assay (ELISA), which was available in kit form (Cozart Bioscience, Abingdon, Oxfordshire, UK). The sensitivity and specificity of this assay were 100% (27). Creatinine was analyzed by the Jaffe reaction (28). All cotinine measurements were expressed per unit creatinine (micrograms per millimole).

Respiratory Control Measurements

Infants were studied between 7:00 P.M. and 7:00 A.M. in an air-conditioned sleep laboratory. On arrival a questionnaire detailing the health of the infant since delivery was completed. Infants were then prepared for study, fed, and allowed to fall asleep naturally. Thereafter, they were fed on demand.

Sleep state was determined from recordings of electroencephalogram (EEG), electrooculogram (EOG), limb movement, and breathing pattern. The signals were recorded and displayed continuously on a computerized polygraph (Sleeplab; Jaeger, Market Harborough, UK) throughout the night. Oxygen saturation was monitored with a Flex II probe on the lateral aspect of the foot and a pulse oximeter (Ohmeda Biox 3700a; BOC, Hatfield, UK). Inspired and end-tidal oxygen levels were monitored at the nostril via a sampling line attached to a mass spectrometer (Airspec 2000; PK Morgan Ltd., Gillingham, Kent, UK). The sampling flow rate was 14.7 ml/min and the sampling time was approximately 290 ms. Ventilation was measured by respiratory inductance plethysmography (RIP) (model 150; Studley Data Systems, Oxford, UK). One RIP band was placed around the chest as close to the axillae as possible and the other was placed around the abdomen at the level of the umbilicus. The mass spectrometer and RIP signals were recorded on a personal computer (Powerpaq Pentium; City Business Systems, Leicester, UK) at a sampling rate of 100 Hz, using custom-built software (RASP; PhysioLogic, Newbury, UK).

During periods of polygraphically confirmed quiet sleep (29) respiratory control was measured by the alternating breath test. A face mask (Rusch size 3; Rendell Baker, Southern Syringe Service Ltd., Southgate, London, UK) through which gases could be delivered was held over the infant's nose and mouth, but did not touch the face. Gas delivery was controlled by computer-operated solenoid valves. The computer was programmed to open the solenoid valve at the start of expiration on the RIP sum signal. This meant that if the gas to be delivered was changed, the pipework and face mask were flushed with the new gas before the infant took the next inspiration. During a baseline period air was delivered from a gas cylinder through the face mask for approximately 2 min. This was followed by a test period during which alternately two breaths of 40% O₂ followed by two breaths of 0% O₂ (100% N₂) were delivered at a flow rate of 6.7 L/min. We aimed to continue this alternating pattern until a continuous test period of 22 breaths was obtained. If the infant sighed the face mask was moved clear of the face until respiration had become regular again, at which point it was repositioned at a point at which 40% O₂ was being delivered. At the end of the test period the inspired gas was then switched back to air. When possible, repeat measurements were made throughout the night during quiet sleep.

Analyses

End-tidal and inspired oxygen concentrations. The inspired (points A and C) and end-tidal (points B and D) oxygen concentrations (percent) were calculated for the second breath of each pair (Figure 1A). In addition, the change in end-tidal oxygen level from the second breath of 40% O₂ (point B) to 0% O₂ (point D) was calculated. The average inspired, end-tidal, and change in end-tidal oxygen levels were then calculated for the whole test.

View larger version (35K): [in this window] [in a new window]   Figure 1.   (A) A short extract of an alternating breath test with a two-breath lag showing points at which 40% O2 and 0% O2 are delivered. The top tracing is the RIP sum signal, in which inspiration is represented by an upward deflection. The bottom tracing is the O2 (%) measured at the nares by the mass spectrometer. Arrows A and C indicate points of measurement of inspired levels of O2 when 40% O2 and 0% O2 are inspired, respectively. Arrows B and D indicate points of measurement of end-tidal levels of O2 when the two gases are inspired. Breaths W, X, and Y are the second breaths of the 40% O2, 0% O2, and the following 40% O2, respectively. These are compared to calculate the percentage alternation. (B) An example of an alternating breath test with a three-breath lag. This example shows how the respiratory response may be underestimated if the conventional analysis method of comparing breaths W, X, and Y is used. The most extreme breaths (indicating maximal response) are represented by W*, X*, and Y*.

Respiratory Control

Runs with a test period of at least 14 continuous breaths and with measurable respired gas signals were analyzed. Sighs (breaths with a tidal volume twice the average tidal volume or greater) were excluded from the analysis.

The RIP sum signal was derived from a weighted combination of the rib and abdomen signals to give an RIP signal proportional to changes in absolute volume. To achieve this a fixed proportionality constant was used (30, 31). Preliminary inspection of our data led us to select a value of K = 0.5. Each breath was identified from the RIP sum signal by a peak-picking routine. Inspiratory and expiratory times (Tr and Te) and semiquantitative tidal volumes in arbitrary units (VTI and VTE) were measured. The following parameters were then derived from these: mean tidal volume (VTT), frequency of breathing (fR), inspiratory drive or flow (VTI/TI), expiratory drive or flow (VTE/TE), inspiratory duty cycle (TI/Ttot), and instantaneous minute ventilation (E).

The analysis of this test has been described previously (26) and is illustrated (Figure 1A). Briefly, for each test period the second breaths of consecutive 40% O₂ and 0% O₂ couplets were compared and the percent alternation for each of the 10 parameters was calculated. The percentage alternation was calculated as the difference between the two breaths expressed as a percentage of the mean. For example, in Figure 1A breaths W and X, and X and Y, will be compared and the percent alternation calculated. For the comparison of breaths W and X the percent alternation is calculated as follows: Percent alternation in VT = {[2 × (VT breath X  VT breath W)]/[(VT breath X + VT breath W)]} × 100. Every second alternation was multiplied by 1, so that the sign of the percent alternation was consistent when a regular alternation occurred. The average percent alternation for each of the 10 parameters was then calculated for each test.

This conventional method of analysis is limited by the fact that only half of the breaths collected are used, since the first breath of each couplet is discarded.

When delivering a stimulus of this nature, there will be a lag between the change in oxygen level in the lung (recorded as end-tidal O₂) and the ventilatory response. The maximal response (taken to be the largest breath after delivery of 0% O₂, or the smallest breath after 40% O₂) may occur two or three breaths after the stimulus. In Figure 1A the smallest breath (X) lags two breaths after the highest end-tidal O₂ (B) and the largest breath (Y) lags two breaths after the lowest end-tidal O₂ (D). Thus in this example the maximal response is seen in breaths that are included in the analysis, because they correspond with the second breaths of each couplet.

In contrast, if the lag in response is equivalent to three breaths, the maximal response breaths occur on the first breath of each couplet and are not used in conventional analysis (Figure 1B). Therefore, in addition to performing the conventional analysis the impact of the lag on comparison of the ventilatory responses in the two groups was estimated for VTI/TI for all tests. For each test the VTI/TI and end-tidal oxygen level were calculated for every breath. These values were then subjected to cross-correlation analysis (MINITAB statistical analysis software) to calculate the lag that gave the best correlation between VTI/TI and end-tidal oxygen level. This approach has been used previously in adults (32). The estimated lag was then used to line up each end-tidal oxygen level with its corresponding response breath and the alternation in VTI/TI for the second breaths was recalculated.

Statistical Analysis

The analysis involved fitting a mixed effects model with normal error structure. A random effect was used to allow for the correlation between repeat runs on the same infant and fixed effects were used for the matching strata and smoking. Significance was assessed by Wald tests (33).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Recruitment

A total of 199 women (97 smokers) were recruited into the study over a 30-mo period. During the time between recruitment and studying the infants 75% of the smoking and 47% of the nonsmoking groups became ineligible for the study. The most likely time for the mother to become ineligible was during the antenatal period, when 36% of the smoking and 22% of the nonsmoking group were excluded, respectively. The single most important reason for becoming excluded from the study was voluntary withdrawal, which was more prevalent in the smoking (31%) than the nonsmoking (15%) group. Considering only those parents who remained willing to bring their infants for overnight study, there were 24 eligible infants from the smoking group, and 54 born to nonsmokers.

Subject Population and Characteristics

Respiratory control measurements were attempted in 35 of the eligible infants born to nonsmoking women and all of the 24 eligible infants in the smoking group. Acceptable data were obtained from 40 infants (17 smoking group), who were divided into 17 closely matched sets, each containing 1 infant from the smoking group and between 1 and 3 infants from the nonsmoking group. The reasons for not obtaining acceptable data from the remaining infants were as follows: (1) infant did not sleep well enough (five infants); (2) technical problems occurred with the mass spectrometer (six infants); (3) a match could not be found (seven infants); and (4) one infant developed a chest infection during the period of the night study, which was subsequently treated with antibiotics. The characteristics of the infants from whom acceptable data were obtained are shown (in Tables 1 and 2). It can be seen that the two groups were similar in many respects. However, the infants in the smoking group were significantly lighter at birth and shorter in length at the time of testing.

The data presented here are the results of a total of 127 tests of respiratory control (56 in the smoking group). The median number of tests for an infant was 3 in both groups, with a range of 1-6 in the nonsmoking and 1-7 in the smoking groups. The average (SD) number of breaths within a test was 32 (12) and 33 (15) in the nonsmoking and smoking groups, respectively.

Cotinine Measurements

The smoking mothers of infants from whom data were obtained reported median cigarette consumption of 10/d (3 to 30). Their median urinary cotinine/creatinine levels at 20 wk of gestation, and in their infants at testing, were significantly higher than those in the nonsmoking group (Table 3).

End-tidal and Inspired Oxygen Measurements

The mean end-tidal and inspired oxygen levels for the two groups of infants are shown (Table 4). The analysis of the differences was carried out on the basis of matched sets. The inspired oxygen levels throughout the test period were not different between the two groups (Table 4). The end-tidal oxygen level was not different between the groups during the baseline period, or when 0% O₂ was delivered during the test period. When 40% O₂ was delivered, the end-tidal oxygen level for the infants in the smoking group was an average of 1.13% higher than that in the nonsmoking group (p = 0.0067) (Table 4). The change in end-tidal levels between the second breaths of each couplet (high end-tidal oxygen  low end-tidal oxygen) was also significantly higher in the smoking group (p = 0.0203) (Table 4).

Respiratory Control Measurements

There was no difference in the baseline breathing parameters measured while breathing air in the 2-min period immediately before the alternations in inspired gases in the two groups. The mean (SEM) respiratory rate was 28.3 breaths/min (1.3), inspiratory time 0.78 s (0.04), expiratory time 1.44 s (0.08) in the smoking group, and 27.3 breaths/min (0.98), 0.81 s (0.03), and 1.49 s (0.07) in the nonsmoking group. Figure 2 presents the mean (± 2 SEM) (adjusted for repeated measures and the matching criteria) for each of the 10 respiratory parameters measured during the test period, using conventional analysis for both groups. It can be seen that variables related to the timing of breathing, TI, TE, fR and inspiratory duty cycle (TI/Ttot), did not alternate as markedly as the parameters related to volume. Figure 3 shows the mean (± 2 SEM) difference in each matched set between the infants in the smoking and nonsmoking groups. Clearly, all the confidence intervals span zero, and therefore there were no significant differences. There was no evidence of a smoking effect in any of the 10 respiratory parameters measured.

View larger version (27K): [in this window] [in a new window]   Figure 2.   The mean (± 2 SEM) alternation for the 10 respiratory parameters measured, adjusted for repeated measures and the matching criteria. The nonsmoking group is represented by the hatched bars, and the smoking group by the stippled bars.

View larger version (12K): [in this window] [in a new window]   Figure 3.   The mean (± 2 SEM) difference within each matched set between infants in smoking and nonsmoking groups for the 10 respiratory parameters.

The proportions of tests with a three-breath lag were similar in the smoking (62%) and nonsmoking (61%) groups. When the alternation in VTI/TI was recalculated to take into account the lag, the mean (SEM) adjusted alternations in VTI/ TI in the smoking and nonsmoking groups were 33.99 ± 3.47 and 35.34 ± 3.12%, respectively. The average difference in adjusted VTI/TI in the matched analysis was 1.35% (p = 0.77).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This carefully controlled matched study has failed to find a significant difference in respiratory responses between infants born to smoking and nonsmoking women. However, we have found unexpectedly that the change in end-tidal oxygen level when 40% O₂ was inspired was significantly greater in the smoking group, despite both groups receiving similar levels of inspired oxygen.

Respiratory Control Measurements

There is evidence from studies looking at the effects of nicotine on the brainstem (19) and carotid body (20) that both pre- and postnatal exposure may affect respiratory responses to hypoxia. However, studies published so far investigating the effect of nicotine exposure on ventilatory responses to hypoxia have been contradictory (21, 23). Lewis and Bosque measured changes in alveolar ventilation in response to hypoxia in the human infant and were unable to find any differences between infants born to smoking and nonsmoking mothers (24). However, this study was performed during a daytime nap, and the authors acknowledge that different results may have been seen during night-time sleep. In addition, this study did not control for potential confounding factors of smoking, and members of the smoking group were significantly older at the time of study. We have found that responses in volume-related parameters increase with postnatal age (34), and therefore the age difference seen in the two groups studied by Lewis and Bosque (24) may have minimized any differences. More recently, it has been reported that infants born to mothers who smoke have blunted ventilatory responses to hypoxia (25). However, these observations were based on a small number of individuals (six in the smoking group, nine in the nonsmoking group) who were studied during daytime sleep after sedation with chloral hydrate. Our aim was to investigate the independent effect of maternal smoking on peripheral chemoresponses to changes in inspired oxygen in unsedated sleeping infants during night-time sleep. Therefore, we conducted a carefully controlled matched study designed to control for confounding factors of smoking so that any independent effect of maternal smoking could be established. We chose the alternating breath test as our method of testing respiratory control, as it has been shown that ventilatory changes in response to this test are mediated peripherally via the carotid chemoreceptors (35). In addition, the experimental procedure used in this study (whereby the inspired oxygen levels alternated between 40 and 0% oxygen) ensured that the mean level of inspired oxygen approximated that of room air so that the infants were not rendered hypoxic, thereby minimizing any central effect of hypoxia.

The ventilatory responses measured in this study were variable both within and between subjects, resulting in wide confidence intervals (Figure 2). The differences between the ventilatory response of the infants in the smoking and nonsmoking groups in each matched set were variable, but showed no evidence of a smoking effect. Therefore, in common with the studies of Bamford and coworkers (23) and Lewis and Bosque (24), we have failed to find an effect of nicotine exposure on respiratory responses to changes in inspired oxygen, suggesting that peripheral chemoresponses are unaffected by nicotine exposure. Evidence has suggested that prenatal nicotine exposure in the rat may lead to changes in central processing that are stimulus and age dependent (36). Our study was not designed to investigate any putative central effects of nicotine, which remain a possibility. In addition, nicotine exposure could influence arousal (24, 37) or cardiac function (38).

The smoking population studied were light-to-medium smokers (median, 10 cigarettes/d) and although there was no evidence of a dose-dependent effect on respiratory control in this study we cannot rule out the possibility that a difference may have been seen with a heavier smoking group.

Our original power calculations were based on preliminary data (26) and indicated that we would need to study 30 infants in each group in order to find a significant difference of 10% in inspiratory drive (VTI/TI) between groups. Although we obtained data from only 17 infants in the smoking group and 23 infants in the nonsmoking group, calculations based on current data indicate that more than 500 infants would have been required to find a significant difference in inspiratory drive between the two groups. Therefore, any effect of smoking is likely to be small and of questionable clinical significance.

Limitations of the Alternating Breath Test

The alternating breath test was first used in infants for measurement of respiratory responsiveness by Blanco and coworkers (39), using two breath alternations of air and 16% O₂. The same group has used breath-by-breath alternations of these gases to investigate development of respiratory responses in the human infant (40, 41). This test was then modified by our group to deliver a greater change in end-tidal levels by supplying two breath alternations of 42% O₂ and 100% N₂ (26). Therefore, the test as applied delivers a four-breath pattern. The response is not instantaneous and there is a lag of about 7 s until maximal response is seen (42). When the test is analyzed breaths 2 and 4 of the pattern are compared, but if the lag was such that the response occurred primarily on breaths 1 and 3, or the lag changed during the test, then the ventilatory response may be underestimated. However, reanalysis of our data, taking account of an estimated lag, did not alter the finding of the study.

End-tidal and Inspired Oxygen Levels

The most striking finding was that infants born to mothers who smoked during pregnancy had a significantly greater end-tidal oxygen level when 40% O₂ was inspired compared with the nonsmoking group. This finding was unexpected and unrelated to inspired oxygen levels. We have no reason to think that gas transfer may differ between the two groups. However, any such difference would be expected to affect end-tidal oxygen levels with both gases, whereas we found no difference in end-tidal oxygen level after inspiration of 0% O₂.

Alternatively, changes in end-tidal oxygen level may be viewed as part of the respiratory response and not purely as a stimulus. As an infant responds to an inspired gas, tidal volume (VT) is changed and therefore the relationship between VT and functional residual capacity (FRC) is altered. If an infant responds to an inspiration of 0% O₂ by increasing VT at a time when 40% O₂ was being delivered, the end-tidal oxygen level would be higher than if a breath of a smaller VT had occurred. Therefore, the nature of the respiratory response affects the end-tidal oxygen levels. It may be that the differences seen in end-tidal oxygen level reflect differences in response in the two groups.

Subject Population

Over the time between recruitment and testing of the infant at around 10 wk of age 77% of the smoking group and 47% of the nonsmoking group became ineligible, or withdrew from the study. A similar study conducted in San Francisco found that 55 and 23% of the smoking and nonsmoking groups, respectively, became unavailable for the study (24). The differences in dropout rates between the two studies are most likely related to the fact that infants in the present study were required to attend the hospital overnight, whereas the measurements reported previously were conducted during the daytime (24). However, it can be seen that dropout rates in studies such as these are appreciable, and that approximately twice as many smoking mothers become ineligible or withdraw compared with nonsmoking mothers. One concern is that this may lead to a biased population being studied. However, there was no difference in gestational age, birthweight, social class distribution, and mode of feeding in the recruited group and those infants whose data were included in the analysis. In addition, the median maternal cotinine/creatinine ratios were similar and the median reported number of cigarettes smoked was 10/d in both groups. Therefore, there was no evidence that the group of infants studied was not representative of the population recruited.

In conclusion, we have failed to find a significant difference in respiratory response to changes in inspired oxygen between infants born to smoking and nonsmoking mothers. However, the change in end-tidal oxygen level when 40% O₂ was inspired was significantly greater in the smoking group. The possible reasons for this include differences in the nature of the response and alveolar ventilation, and warrant further investigation.