Main

SIDS is considered a heterogeneous clinical entity(1). Compared with control subjects, some future SIDS infants were characterized by repeated upper airway obstructions during sleep(2–4) and by symptoms attributed to ANS dysfunctions, such as tachycardia(5), bradycardia(6), higher overall HR and smaller HR variability(7–9), prolonged Q-Tc index(10,11), or excessive sweating(1,2,12). It was suggested that infants who subsequently died of SIDS suffered from both impaired vagal tone and increased sympathetic activity(5,7–9,13,14). Some authors have pointed out the importance of an intact ANS to elicit HR changes after obstructive events(15–17). The purpose of the present study was to evaluate autonomic cardiac responses to obstructive apneas in future SIDS victims and in control infants by means of autoregressive spectral analysis of the infants' HR.

METHODS

Patients. Between May 1992 and May 1995, a polygraphic sleep study was performed on 18 infants who later became victims of SIDS. These recordings were collected from 13 sleep laboratories. The sleep studies were conducted as part of different sleep research programs or to alleviate parental anxiety about sleep apnea. The present study was approved by the Institutional Review Board, and parents gave their informed consent. Of the 18 SIDS, 13 infants were males and six were preterm infants. Two preterm infants had inappropriate weight for gestational age. One infant was a sibling of a SIDS victim. No infant was being monitored at the time of death. The 18 deaths were unexpected and remained unexplained despite postmortem studies. The recording of two control infants was selected for each SIDS recording. Control and SIDS infants were matched for sex, gestational age, postnatal age, weight at birth, and sleeping position. All control infants were asymptomatic, had no family history of SIDS, and survived the first year of life uneventfully. At the time of recording, no infant had signs of infection and none were receiving medication. Data on the children's history and usual behavior were collected using a standard questionnaire before sleep monitoring was undertaken. The questionnaires were coded and analyzed with the sleep recordings.

Polygraphic recordings. The infants were admitted for a night monitoring session (8-9 h). The data were collected on a computerized polygraph recording system (Morpheus system, Medatec, Belgium). The following variables were recorded simultaneously: two EEG, two electrooculograms, digastric electromyography, ECG (DII), thoracic and abdominal respiratory movements by inductive plethysmography, and airflow by means of thermistors taped under each nostril and on the side of the mouth. Oxygen saturation was recorded continuously using a transcutaneous sensor (Ohmeda Box, USA).

Data analysis. Each record was allocated a random code number. The code was disclosed after completion of the analysis. Two independent scorers analyzed the sleep recordings to ensure reliability. The two scorers had not taken part in the collection of the data and analyzed the coded recordings without knowledge of the patient's identity. Disagreements were discussed and subsequently agreed upon codes were used in data analysis. Thirty-second periods of the recordings were analyzed and categorized as NREM sleep, REM sleep, indeterminate sleep, or wakefulness according to criteria recommended in the literature(18,19). NREM refers to NREM II and III stages. Behavioral arousal was defined as opening the eyes. Gross body movements were measured by actigrams and confirmed visually. Sleep efficiency was defined as the ratio of the total sleep time divided by the total recording time, expressed as a percentage. Mean values for oxygen saturation, HR, and respiratory rate were calculated on 1-min stable sleep periods, at least 15 min after any change in body position, movement, sigh, or arousal artifact. Baseline HR and saturation in oxygen were calculated during 1-min periods preceding respiratory events. Transient arousals after obstructive events were defined through the presence of EEG changes associated with irregular respiratory and body movement(20). Sleep apneas were scored only if they lasted 3 s or more. A central apnea was scored when flat tracings were obtained simultaneously from strain gauges and thermistors. Periodic breathing was defined by the succession of at least three central apneas separated from each other by <20 s of breathing. An obstructive apnea was scored when continuous deflections were obtained from strain gauges while a flat tracing was recorded from thermistors. To avoid artificial scoring due to thermistor displacement, obstructive apneas preceded by body movements, crying, or sighs were rejected. Mixed apneas were defined as central apneas followed directly by obstructive episodes and were scored with obstructive apneas.

Selection of obstructive apneas. Two obstructive events were selected in the recordings of each infant. Obstructive apneas were selected with similar characteristics in the two populations: time of night, sleep stage, duration, prior apnea oxygen saturation, postapnea desaturation, and postapnea HR deceleration.

HRSA. Digitized ECG signals were sampled at 300 Hz. Periods of 256 successive RR intervals were selected preceding the event and after the end of the obstructive apnea (Figs. 1–3). Premature ventricular contractions or artifactual RR intervals due to gross body movements or arousals were eliminated by visual analysis of the HR data before HRSA was performed. A HRSA of the trendgram was calculated for each period, and center frequency and power of each spectral component were obtained(21,22) (Figs. 2 and 3). Two major peaks were recognizable: a LF component defined by center frequency of 0.1 Hz Eq (0.04-0.15 Hz Eq) and a HF component defined by a center frequency of 0.4 Hz Eq (>0.15-2 Hz Eq)(21). Respiratory frequency during the selected period was measured manually after being printed on paper. For each 256-RR interval period, the major component in the LF band of the HR spectrum was related to the major component in the HF band corresponding to the mean respiratory frequency as determined by analysis of breathing rate. The ratio of LF/HF powers for each episode was calculated as an index of sympathovagal interaction(23). Because of the possibility of variation of the breathing rate within the selected segment, we compared the normalized powers obtained by this method (major component of the band) with the results obtained when the totality of the band was studied: LF band, 0.04-0.15 Hz Eq; HF band, >0.15-2 Hz Eq. Spectral components were represented as RR intervals (in ms), power (in ms2), bandwidth (in Hz Eq), and normalized power values obtained by dividing the power of the period by the total power component (in percent) after subtraction of the direct current component(21,22). The optimal autoregressive model order was determined by minimizing the value of the final predictor error(24). Stationary was confirmed by a pole diagram analysis(21,22).

Figure 1
figure 1

Polysomnographic recording of a mixed apnea. From surface ECG, time series of RR intervals were calculated as a function of beats before (2) and after the end (3) of the mixed apnea. EMG, electromyogram; EOG1, 2, electrooculogram 1, 2; F3-C3 and F4-C4, bipolar EEG; VTH and VAB, thoracic and abdominal volumes; NAF, nasal thermistor; SAO2, oxygen saturation.

Full size image
Figure 3
figure 3

(A) RR interval trendgram after apnea. (B) Autoregressive power spectrum of 3.

Full size image
Figure 2
figure 2

(A) RR interval trendgram before apnea. (B) Autoregressive power spectrum of 2. c/b, cycles per beat.

Full size image

Statistical evaluation. The Friedman test was applied to compare the future SIDS victim and the two control subjects. When the test was significant, Wilcoxon matched-pairs signed rank test was used to compare the SIDS case with each of the matched control subjects and to compare the two control subjects with each other. The test was considered positive when significant differences were observed between the SIDS case and the two control subjects, but not between the two control subjects(25). Because of the fewer number of obstructive events in the control infants, Wilcoxon nonpaired matched test was applied to compare the spectral analysis values between the future SIDS victims and the control group. The medians presented in the tables are those obtained by pooling the two control subjects. The χ2 test was used with 2 × 2 tables.

RESULTS

General characteristics of the population. Due to the study design, no differences existed between the future SIDS and the control subjects for the following variables: sex, gestational age, postnatal age, weight at birth, and sleeping position. Mothers of future SIDS victims were younger [median 24 y for SIDS infants and 29 y in control subjects; range 20-43 y (p = 0.038)] and more frequently smoked during pregnancy [4/18 for SIDS infants and 3/36 for control infants (p = 0.003)] (Table 1).

Table 1 Major characteristics of infants studied
Full size table

Sleep characteristics. After sleep recordings, no significant differences were seen between the two groups of infants for the following variables: total recording time (median duration 480 min; range 330 to 510 min), total sleep time (median 347 min for SIDS infants, 370 min for control infants; range 201 to 478 min), sleep efficiency (median 80.2% for SIDS infants, 77.25% for control infants; range 47.5-100%), NREM sleep (median 29.95% for SIDS infants, 32% for control infants; range 16.6-45.5%), REM sleep (median 46.05% for SIDS infants, 42% for control infants; range 22.7-69.8%), sleep stage changes (median 15.5 for SIDS infants, 17 for control infants; range 11-23), movements (median 4.65% for SIDS infants, 3.75% for control infants; range 0-17.4%), arousals (median 15.4% for SIDS infants, 17.8% for control infants; range 0-49.9), frequency of central apneas per infant (median 27 for SIDS infants, 32.5 for control infants; range 3-106), maximum duration of central apneas (median 8 s for SIDS infants, 7 s for control infants; range 4.1-18.9 s), basal HR in NREM sleep [median 133 beats per minute (bpm) for SIDS infants, 127.3 bpm for control infants; range 110-155 bpm], basal HR in REM sleep (median 136 bpm for SIDS infants, 129 bpm for control infants; range 111-156 bpm), basal breathing rate in NREM sleep (median 39 breaths per minute for SIDS infants, 31 breaths per minute for control infants; range 22-62 breaths per minute), basal breathing rate in REM sleep (median 43 breaths per minute for SIDS infants, 36 breaths per minute for control infants; range 25-62 breaths per minute), and oxygen saturation values (median 93.7% for SIDS infants, 95.2% for control infants; range 86.5-98.6%). After apneas, no differences were seen for the percentage of HR deceleration associated with central or obstructive events or for the percentage of oxygen desaturation associated with both types of apneas.

Characteristics of obstructive events. As shown in Table 2, more future SIDS than control infants had obstructive or mixed apneas. Each future SIDS victim had more obstructive and mixed episodes than their matched control subjects. No significant differences were found between the two groups for the duration of the obstructive or mixed apneas. Obstructive events occurred mainly during REM sleep in the two populations (84.5% in future SIDS victims and 95.8% in control infants; p = NS). All obstructive events were thus selected during REM sleep.

Table 2 Sleep and cardiorespiratory characteristics that statistically differentiated SIDS cases from matched control infants
Full size table

Two obstructive events were selected in each SIDS infant with obstructive events (34 events) and, according to the number of obstructive events, zero to two obstructive events were selected in each control subject (37 events). There were 16 control infants and one future SIDS infant without obstructive event (p = 0.009), three control infants and no future SIDS infant with one obstructive event (p = NS), and 17 control infants and 17 future SIDS infants with at least two obstructive events (p = 0.002). No difference was seen between the two groups for the duration of obstructive events (median 5.2 s in future SIDS victims and 5.1 s in control infants; range 3.2-12.1 s) or oxygen saturation values before obstructive events (median 98%; range 93-100%). After the obstructive apneas, no differences were seen for oxygen saturation values (median 94.5% in future SIDS victims and 96% in control infants; range 82-99%), the percentage of oxygen desaturation (median 2.5% in future SIDS victims and 3% in control infants; range 3-15%), the HR values (median 120 bpm for future SIDS victims and 112 bpm for control group; range 60-153 bpm), or the percentage of HR decelerations (median 23.5% for SIDS infants and 30.5% for control infants; range 9-81%). No differences were noted in the frequency of transient arousals after obstructive events (26/34 in future SIDS victims and 29/37 in the control group). There was no difference in the ratio of obstructive to mixed apneas in the two populations (20/14 for future SIDS infants and 19/16 for control infants). There was no difference in the nycterohemeral selection of obstructive events (median 0511 h for SIDS infants and 0555 h for control infants; range 0003-2357 h). The obstructive events in REM sleep were divided according to duration (3 to <5 s, 5 to <10 s, and >10 s). There were 15 apneas of <5 s in SIDS infants and 18 in control infants, 19 apneas of 5 to 10 s in SIDS infants and 17 in control infants, no apnea of >10 s in SIDS infants and two in control infants (NS).

Comparison of HRSA between SIDS and control infants. When both populations were compared before and after the obstructive apneas by means of HRSA (Table 3), HF powers and HF normalized powers were significantly lower, and the LF/HF power ratios higher in future SIDS victims than in control subjects. After the obstructive apneas, these differences between the two populations were greater than in the preapnea situation. HF bandwidth was larger in the future SIDS victims than in the control infants before obstructive apneas. Comparing the HRSA, no difference was seen between the two groups for total spectrum power, RR intervals, normalized or nonnormalized LF powers, and LF bandwidth between the two populations, either before or after the obstructive sleep events. The results were similar when the major components within the band or the totality of the spectral band were studied.

Table 3 HRSA before and after obstructive apneas
Full size table

Comparison of HRSA before and after obstructive events. Comparing HR power values for apnea-associated changes, significantly lower LF/HF power ratios were measured after obstructive apneas in control infants, whereas no difference was seen in future SIDS victims (Table 3). Of the 37 obstructive episodes studied in control infants, 29 were associated with a decrease in LF/HF energy and eight with an increase. Of the 34 obstructive episodes studied in the future SIDS group, 18 were followed by a decrease and 16 by an increase in LF/HF power (p = 0.044). After the apneas, there was a significant decrease in LF normalized power in the two populations. Total spectrum power increased significantly as well as HF power in control subjects after obstructive apneas. These differences were not found in future SIDS victims. When the obstructive events were partitioned according to duration (Table 4), these differences were found mainly in the long apneas. No differences were seen when the changes in HRSA associated with obstructive events were studied in relation to sleep position, sex, prematurity, or maternal smoking habit.

Table 4 HRSA according to apnea duration
Full size table

DISCUSSION

The obstructive apneas seen during REM sleep were followed by significantly more frequent decreases in LF/HF ratio in the control than in the future SIDS infants. The difference in autonomic reactions cannot be attributed to experimental conditions such as age, body position during sleep(26), time of night, feeding times, characteristics of obstructive events (duration, time, saturation, or HR changes), or sleep fragmentation(27).

We must admit several limitations to the present study. First, we hypothesized that before and after obstructive apneas, the HR varies in a sinusoidal fashion, a prerequisite for the application of Fourier analysis. For this reason, an autoregressive spectral analysis method was used because it was well suited for the analysis of short time series(28). Second, no spectral analysis was performed on respiratory movements, and cross-spectral analysis of respiration and HR changes were not evaluated(28). We studied the major component in the HF band corresponding to the mean respiratory frequency and the HF band to consider the breathing variations within the selected segment. When comparing future SIDS victims with control infants, HF energy was seen to decrease during REM sleep. These results are similar to those reported by Kluge(14) with the use of cross-spectral analysis of respiration and HR changes. Third, the interpretation of the LF/HF ratio is complex. Within the LF range, the HR fluctuations depend on both symphathetic and parasympathetic controls. The band under 0.09 Hz represents fluctuations of vasomotor or thermal origin(29); the mild-frequency band (between 0.1 and 15 Hz) has been related to baroreceptor control(30). The power in these two frequency bands usually merges into one wide peak. The respiratory peak has been shown to be mainly vagally mediated(31). The vagal efferent fibers could not originate in a common brainstem structure. The HF peak that corresponds to the respiratory sinus arrhythmia would reflect only the part of the vagal efferent system from the nucleus ambiguus(32). Spectral HR techniques do not permit evaluation of the influence of the other branch of the parasympathetic system from the dorsal motor nucleus(32), although it was suggested that competition between the two branches of the parasympathetic system could be responsible for autonomic dysfunction(32,33).

With these restrictions in mind, the present study highlights that after OSA, future SIDS victims do not exhibit the decrease in LF/HF ratio seen in control infants.

In OSA adult patients, sleep apneas are associated with marked hemodynamic oscillations consisting of bradycardia during apneas, followed by a transient increase in HR and arterial pressure with resumption of breathing(34–37). In OSA adults, muscle sympathetic nerve activity recorded from peroneal nerve revealed that apneas were accompanied by an increase in sympathetic activity that resolved on resumption of breathing(34,36). This postapnea decrease in sympathetic activity was attributed to lung inflation and to changes in arterial baroreflex activity(38).

The cardiac responses observed in the control infants could thus result from various changes in autonomic controls, such as decreases in orthosympathetic tone, or increases in parasympathetic activity or a combination of both. The lack of decrease in LF/HF ratio measured in the future SIDS victims could result from an impairment of ANS control, as postulated in these infants(5,7–9,13,39). The future SIDS victims had lower values of normalized and nonnormalized HF powers and higher values for the LF/HF power ratios than the control subjects. Similar findings were reported previously(14,40,41). Such potential HR dysautonomia could reduce the electrical stability of the heart(7,39) and precipitate ventricular fibrillation and sudden cardiac death(42,43). An intact adrenal function is necessary for gasping and surviving from anoxia in newborn rats(44). Such response is blunted in growth-retarded newborn rats with high basal sympathoadrenal activity(45). An attenuated vagal or an increased sympathetic activity could reduce behavioral adaptation to environmental stresses(33,46). An autonomic dysautonomia could explain the inhibition of some homeostatic reflexes, as reported in some future SIDS infants, including a deficiency in ventilatory and arousal responsiveness to hypercarbia or hypoxia(47,48), a decrease in consistent sustained HR changes(9), or an increased frequency of obstructive events(3,4).

CONCLUSION

Compared with control infants, future SIDS victims were characterized by different autonomic status and autonomic responses to obstructive apneas during sleep. It could be hypothesized that these characteristics may be associated with a higher vulnerability to external or endogenous stress factors.