- Top of page
The efficacy of the nasobuccal thermistor (NT) was compared with the nasal pressure cannula (NC) and the calibrated, time-differentiated respiratory inductance plethysmography sum signal (DS) in the detection of obstructive events in children during polysomnography (PSG). The overnight PSG of 20 consecutive referrals were selected for analysis. Obstructive events were scored in each study three times by one operator using a blinded procedure whereby either the NT, the NC or the DS was visible. The standard PSG channels were also visible. SPSS software was used for statistical analysis. Twenty patients aged 5 weeks to 16 years were studied. Agreement in obstructive apnoea–hypopnoea index (OAHI) was highest between the NT and NC, and the NC and DS. The NC signal was significantly more likely to be uninterpretable than the NT (P = 0.02) and this did not correlate with age. Event detection by the NT was significantly improved by the addition of either the NC (P = 0.01) or the DS (P = 0.001), while the NC stood alone unless the DS was added (P = 0.02). There was no significant difference in OAHI by the NC versus the DS. The NC detected significantly more OA than the NT or the DS (P = 0.04), while the DS trended towards detecting more OH. There was no significant difference in OAHI between any combination pair. The nasal cannula and differentiated sum signal perform better as measures of paediatric airflow than the NT. To optimize the detection of obstructive events in children we recommend using at least one, if not both these methods in paediatric sleep laboratories.
- Top of page
Obstructive sleep apnoea (OSA) syndrome in children is diagnosed by overnight polysomnography and affects up to 2% of the population (Ali et al., 1993). Measurement of airflow is fundamental to the identification and quantification of obstructive events during sleep. The reference standard for direct airflow measurement is the pneumotachograph, however it requires a tight fitting mask which can be detrimental to sleep and is impractical for use in routine polysomnography. Hence most paediatric laboratories use indirect measures to estimate airflow limitation.
A nasobuccal thermistor (NT) detects temperature changes between inspired and expired air and has been the surrogate method recommended by the American Thoracic Society Guidelines (1996) for airflow measurement in paediatric laboratories. There are, however, significant limitations to this measurement. First, the signal amplitude does not accurately reflect the actual airflow magnitude because of the long time-constant response (Norman et al., 1997). This particularly adversely affects the detection of obstructive hypopnoeas (OH) which are the commonest respiratory event in childhood OSA. Secondly, this approach does not allow analysis of the flow/time contour (Hosselet et al., 1998).
Nasal cannula pressure measurement (NC) is an alternative non-invasive correlate of airflow that has been validated in adults and found to have excellent agreement compared with a pneumotachograph (Heitman et al., 2002; Thurnheer et al., 2001). The NC has been found to detect 30% more obstructive respiratory events than the NT in adults (Norman et al., 1997) and up to 58% more in children (Trang et al., 2002). Furthermore, it has recently been shown that in adults with OSA analysis of the flow signal from a nasal cannula can provide a respiratory disturbance index (RDI) similar to that obtained in a full polysomnography (Ayappa et al., 2004).
Quantitative respiratory inductance plethysmography (RIP) measurements detect changes in abdominal and thoracic volume. The calibrated sum signal of the RIP waveform has been shown to be accurate in the analysis of tidal volume when compared with a whole body plethysmograph and a pneumotachograph (Carry et al., 1997). This volume signal can be differentiated by time, and the resultant differentiated sum signal has been used as an indirect measure of airflow in adults (Kaplan et al., 2000; Loube et al., 1999). This method has not yet been tested in children.
We sought to compare the efficacy of the NC and the differentiated RIP sum signal (DS) with the routinely used NT in the detection of obstructive apnoeas (OA) and OH in children.
- Top of page
Twenty patients aged 5 weeks to 16 years were studied. Table 1 shows the mean and standard deviation, as well as median and range for age, total sleep time (TST), OAHI and percentage time with uninterpretable signals of each method (NC-UI, NT-UI and DS-UI).
Table 1. Descriptive statistics
Correlation was significant between all three methods for OA, OH and OAHI, but tightest between NC and DS (r = 0.97, P = 0.01). Agreement between the individual methods for OAHI is demonstrated by the Bland–Altmann plots in Fig. 1. There was high agreement between the NC and NT, and the NC and DS. The agreement in OAHI between the DS and NT was poorer (Fig. 1c– mean difference significantly different to zero. Represented in Table 2, row 2, where OAHI by DS is significantly higher than by NT, P = 0.02).
Figure 1. Bland–Altmann plots for OAHI. (a) Mean difference between NC and NT not significantly different from zero. (b) Mean difference between NC and DS not significantly different from zero. (c) Mean difference between DS and NT significantly different from zero (P = 0.02).
Download figure to PowerPoint
Table 2. Comparison of methods by OAHI (median value, interquartile range (P25–P75) and P-value for Wilcoxon signed ranks tests assessing all method combinations)
|Method 1||Method 2||Method 1||Method 2||P-value|
|NT||NT + NC||0.8||0.2||10.5||2.0||0.4||10.3||0.01|
|NT||NT + DS||0.8||0.2||10.5||1.0||0.2||12.6||0.001|
|NT||NC + DS||0.8||0.2||10.5||2.0||0.4||12.5||0.02|
|NC||NT + NC||3.0||0.2||12.8||2.0||0.4||10.3||0.33|
|NC||NT + DS||3.0||0.2||12.8||1.0||0.2||12.6||0.20|
|NC||NC + DS||3.0||0.2||12.8||2.0||0.4||12.5||0.02|
|DS||NC + DS||1.0||0.2||12.6||2.0||0.4||12.5||0.06|
|DS||NT + NC||1.0||0.2||12.6||2.0||0.4||10.3||0.62|
|DS||NT + DS||1.0||0.2||12.6||1.5||0.2||12.5||0.02|
|NC + NT||NT + DS||2.0||0.4||10.3||1.5||0.2||12.5||0.33|
|NT + DS||DS + NC||1.5||0.2||12.5||2.0||0.4||12.5||0.26|
|DS + NC||NC + NT||2.0||0.4||12.5||2.0||0.4||10.3||0.41|
The NC signal was significantly more likely to be uninterpretable than the NT (P = 0.02), but not the DS when an assessment of uninterpretable signals for >20% of total sleep time was made. The time spent with an uninterpretable signal did not correlate significantly with age by any of the three methods. One signal being uninterpretable for between 20% and 60% of the study contributed to outliers in the Bland–Altmann plots in three of six instances.
The three methods were compared individually, and then each single method against all other combination pairs. Finally a pair-by-pair method comparison was made. Table 2 shows the median, 25th and 75th percentile values (Wilcoxon signed ranks test for non-parametric data) for OAHI when a comparison of single and combined methods was made. The DS alone detected significantly more events than the NT alone (P = 0.02), but not the NC. Addition of the NC or the DS to the NT significantly increased event detection (P = 0.01, 0.001), as did a combination of NC and DS compared with the NT alone (P = 0.02). Addition of only the DS and not the NT to the NC increased event detection significantly (P = 0.02). Addition of only the NT to the DS increased event detection significantly (P = 0.02), although adding the NC to the DS trended towards significance (P = 0.06). When the ability of all combination pairs for OAHI was examined, there was no significant difference between any combination pair.
Further analysis of individual events (OA and OH, rather than OAHI) showed that the NC detected OA better than both the NT and DS (P = 0.04; row 3, Table 3). Analysis of paired methods showed that the NT and NC combination was significantly better at detecting OA than the NT and DS combination (P = 0.04; row 7, Table 3). The NT frequently missed the events, and the DS tended to label them OH instead of OA. Hence, the DS trended towards detecting more OH than the NT (P = 0.06) and the NC (P = 0.07) (row 12, Table 3). The combination of NC and NT measured more events than the individual methods alone (rows 13 and 14, Table 3, P = 0.02 and 0.03, respectively). Adding the DS to the NT improved OH detection more than adding the NC to the NT (P = 0.05; row 15, Table 3), hence the DS/NT pair was superior for OH detection.
Table 3. Comparison of methods by OA and OH
|Method 1||Method 2||Method 1||Method 2||P-value|
|NT||NT + NC||0.0||0.0||22.7||0.0||0.0||10.0||0.08|
|NC||NT + NC||0.0||0.0||43.1||0.0||0.0||10.0||0.50|
|DS||NT + DS||0.0||0.0||16.2||0.0||0.0||6.0||0.31|
|NC + NT||NT + DS||0.0||0.0||10.0||0.0||0.0||6.0||0.04|
|NT + DS||DS + NC||0.0||0.0||6.0||0.0||0.0||7.0||0.78|
|DS + NC||NC + NT||0.0||0.0||7.0||0.0||0.0||10.0||0.11|
|NT||NT + NC||6.0||0.9||55.0||13.0||1.0||57.0||0.02|
|NC||NT + NC||9.7||0.9||56.6||13.0||1.0||57.0||0.03|
|DS||NT + DS||10.0||0.9||62.0||10.0||1.0||64.0||0.13|
|NC + NT||NT + DS||13.0||1.0||57.0||10.0||1.0||64.0||0.05|
|NT + DS||DS + NC||10.0||1.0||64.0||13.0||1.0||60.0||0.50|
|DS + NC||NC + NT||13.0||1.0||60.0||13.0||1.0||57.0||0.12|
- Top of page
In this study we have demonstrated that both NC and DS can be used as surrogate airflow measures in children to detect OA and OH. This is the first report of the use of the differentiated sum signal in paediatric sleep medicine.
We have also shown that both the NC and the DS function better alone than the NT. The NC performs significantly better in detection of OA than the other two methods, while the differentiated sum was better for OH detection. Adding either the NC or the DS to the NT always improved the OAHI. Conversely, the OAHI by the NC alone could only be improved by the addition of the DS. The DS alone performed no differently from a combination of the NC and NT, but the OAHI could be significantly improved by adding the NC. Finally, when any two methods were used concurrently, there was no significant difference in the OAHI. However, based upon the above observations of individual events, the combination of the NC and the DS seems desirable. These findings have considerable implications for the diagnosis of OSA in paediatric sleep units where the thermistor is still considered the standard airflow measure.
The differentiated sum of the RIP signal identified more OH but less OA than the NC or the NT. At least some of the OH were identified as OA by the NC. One reason for this may relate to the fact that the DS method assumes a closed system between the thorax and abdomen. In some instances a contribution from the anatomical dead space may lead to incomplete flattening of the flow contour during complete obstruction. The DS has also been validated against oesophageal pressure in the detection of airflow limitation in adults (Kaplan et al., 2000; Loube et al., 1999) using various computer-derived algorithms to classify flow limitation. In contrast we used simple visual analysis of the shape of the flow contour to assist in identification of obstructive events which may introduce an observer bias and account for some variation in event labelling. In routine clinical practice, however, it may not be necessary to distinguish OA from OH as both events have similar pathophysiology (American Academy of Sleep Medicine, 1999), and this distinction is unlikely to affect treatment.
One drawback to the use of the NC is the potential for unreliable signals and this has been highlighted previously as a limiting factor (Trang et al., 2002). In our study, the time spent with an uninterpretable NC signal was longer than has previously been reported (median 22.7% of TST, range 0–100%). The proportion of patients with uninterpretable signals for more than 20% of TST was significantly higher for the NC than the NT (P = 0.02). An uninterpretable nasal cannula signal between 20% and 60% of sleep time explained half the outliers in the Bland–Altmann plots. Uninterpretable signals related most often to mouth-breathing which is a relatively common occurrence in childhood OSA. Some authors have argued that this limitation is counterbalanced by the efficacy of this technique to detect obstructive events (Farre et al., 2001). Interestingly two of our outliers in the Bland–Altmann plots favoured NC efficacy and were associated with full interpretability, suggesting this study may have underestimated the efficacy of the NC due to the high percentage of complete mouth-breathers. Although NC efficacy was again demonstrated by our study, it was not reliable enough to be used as the only measurement of airflow and we would suggest using a backup method. The DS is an appropriate first choice given that signal reliability (median 4% of TST uninterpretable, range 0–100%) was not significantly worse than the NT and event detection was significantly better (P = 0.02).
Trang et al. (2002) have previously validated nasal pressure events in children using oesophageal pressure studies, discounting the possibility of false-positive event detection. We did not compare events detected by each of the three methods with swings in oesophageal pressure because this method is not routine in our sleep unit. This comparison would have clarified that we were not overdiagnosing events, particularly with the NC and the DS. We did, however, only score hypopnoeas if there was a 50% or more reduction in airflow accompanied by a 3% desaturation or an arousal, suggesting that these were in fact true events. This is a conservative method compared with the American Academy of Sleep Medicine criteria which suggest events can be scored either with a 50% reduction in airflow, or with less reduction in flow but the presence of a 3% desaturation or an arousal (American Academy of Sleep Medicine, 1999). In addition, the relationship between nasal pressure and actual airflow is nonlinear, which in itself may result in an underestimation of airflow and reduce event detection (Heitman et al., 2002). Furthermore, it has been recently suggested from assessment of milder flow limitation on nasal pressure traces that clinical complaints and common breathing abnormalities in children are being overlooked by published polysomnographic scoring recommendations (Guilleminault et al., 2004).
In conclusion, both the nasal cannula and differentiated sum signal perform significantly better in the detection of obstructive events than the NT which is the current paediatric standard airflow measure. If only one flow measure can be used, given the problems with nasal cannula reliability, we would advocate the use of the differentiated sum signal. Where possible, however, we would recommend that more than one airflow method be used routinely in paediatric sleep laboratories, and our preference is the combination of the nasal cannula and differentiated sum signal.