Tidal breathing flow profiles during sleep in wheezing children measured by impedance pneumography

TITLE: Tidal breathing flow profiles during sleep in wheezing children measured by1 impedance pneumography2 AUTHORS: Gracia-Tabuenca, Javiera; Seppä, Ville-Pekkab; Jauhiainen, Millaa; Paassilta, Maritac;3 Viik, Jaria; Karjalainen, Jussic 4 5 AFFILIATIONS: a Faculty of Medicine and Health Technology, Tampere University, Korkeakoulunkatu6 10, FI-33720, Tampere, Finland7 b Revenio Research Ltd., Äyritie 22, FI-01510, Vantaa, Finland8 c Allergy Centre, Tampere University Hospital, Teiskontie 35 PL 2000, FI-33521,9 Tampere, Finland (primary study place)10 11 *CORRESPONDING AUTHOR Javier Gracia-Tabuenca12 Email: [email protected] Phone: +35846908698114 Address: SM319, Korkeakoulunkatu 10,15 Zip: FI-3372016 City: Tampere17 Country: Finland18 19 CONFLICTS OF INTEREST: V.-P. Seppä and J. Viik are shareholders in Tide Medical Oy, which holds20 patents related to impedance pneumography. V.-P. Seppä is an employee of Revenio Group Oyj, which21 commercializes impedance pneumography technology.22 23 FUNDING: This work was supported by the Tampere Tuberculosis Foundation and the Tampere24 University of Technology Graduate School.25 26 27 28 This is the accepted manuscript of the article, which has been published in Respiratory Physiology and Neurobiology. 2020, 271, 103312. https://doi.org/10.1016/j.resp.2019.103312 ABSTRACT1 For the first time, impedance pneumography (IP) enables a continuous analysis of the tidal breathing2 flow volume (TBFV), overnight. We studied how corticosteroid inhalation treatments, sleep stage, and3 time from sleep onset modify the nocturnal TBFV profiles of children. Seventy children, 1-5 years old4 and with recurrent wheezing, underwent three, full-night TBFVs recordings at home, using IP. The first5 recorded one week before ending a 3-months inhaled corticosteroids treatment, and remaining two,6 2 and 4 weeks after treatment. TBFV profiles were grouped by hour from sleep onset and estimated7 sleep stage. Compared with on-medication, the off-medication profiles showed lower volume at8 exhalation peak flow, earlier interruption of expiration, and less convex middle expiration. The9 differences in the first two features were significant during non-rapid eye movement (NREM), and the10 differences in the third were more prominent during REM after 4 h of sleep. These combinations of11 TBFV features, sleep phase, and sleep time potentially indicate airflow limitation in young children.12 13 14 KEYWORDS: lung function, tidal breathing, wheezing children, impedance pneumography15 16 1. INTRODUCTION1 Tidal breathing flow volume (TBFV) analysis has been proposed as an alternative for detecting lower-2 airway obstruction in young children who are unable to cooperate with forced spirometry (Beydon et3 al., 2007). However, interpretation of TBFV profiles, particularly the expiratory limb, is challenging4 (Bates, 1998). Expiration is shaped not only by passive mechanical characteristics such as5 thoracopulmonary recoil and airway resistance (Otis et al., 1950), but also by the active braking during6 the early part of expiration and the active interruption of the expiration ending (Hutten et al., 2008).7 Airway narrowing directly alters passive characteristics, which triggers subject-dependent active8 adaptation strategies (Baldwin et al., 2006; Maarsingh et al., 2000; Morris and Lane, 1981). Moreover,9 passive and active characteristics are influenced by other factors such as instrumentation (Fleming et10 al., 1982), awareness state (Lodrup-Carlsen and Carlsen, 1993), and other respiratory conditions11 (Leonhardt et al., 2010).12 Impedance pneumography (IP) allows the continuous recording of the tidal flow overnight, providing13 several advantages to TBFV profile analysis. Firstly, IP uses four surface skin electrodes to derive the14 respiratory flow noninvasively, from changes in the thoracic electrical impedance, which is15 proportional to lung aeration. Hence, unlike conventional pneumography (PNT) (Fleming et al., 1982),16 IP does not corrupt the shape of the TBFV profiles. Secondly, sleep is characterised by a decrease in17 respiratory musculature tone, which is accentuated further during the rapid eye movement (REM)18 stage (Horner, 2010). It has been hypothesized that a decrease in muscle tone increases the19 contribution of passive characteristics to the TBFV profile, revealing signs of obstruction (Gracia-20 Tabuenca et al., 2019). Thirdly, diseases such as asthma are influenced by many neural, hormonal, and21 autoimmune circadian factors. Asthma symptoms worsen late at night and early in the morning22 (Bohadana et al., 2002). Assessment of the nocturnal TBFV profiles recorded at home may reveal23 symptoms that go unnoticed by tests conducted in hospitals.24 Previous studies have proven the feasibility of using IP to derive TBFV profiles during night sleep25 (Gracia-Tabuenca et al., 2019; Seppä et al., 2016) and the strong agreement between IP and PNT26 results in children (Seppä et al., 2013b) and infants (Malmberg et al., 2017), even under induced27 bronchoconstriction. The effect of interactions between asthma risk and sleep physiology on the28 shape of TBFV profiles has been studied for infants (Gracia-Tabuenca et al., 2019), but not for older29 children. During the first year of life, development of the thoracic cage (Allen and Gripp, 2002) and30 the nervous system (Rabbette et al., 1994) modifies breathing strategy. For example, dynamic31 maintenance of end-expiratory volume decreases (Colin et al., 1989), thoracoabdominal asynchrony32 (Guilleminault et al., 1982) decreases, and respiratory apnoeas become less common (Flores-Guevara33 et al., 1982). It is unknown if these developmental changes affect the results observed in infants.1 Finally, lower airway obstruction increases with night progression, at least in adults with and without2 asthma (Bellia et al., 1989). However, whether TBFV profiles change with night progression has not3 been studied.4 The present longitudinal study assessed the effect that the interruption of medication had on the TBFV5 profiles obtained from overnight IP recordings taken at home, for a group of wheezing children. It also6 investigated the extent to which the time from sleep onset, as well as REM and non-REM sleep stages,7 influenced these changes. These two main sleep stages were respectively estimated from regions of8 high and low respiration variability.9 10 2. MATERIALS AND METHODS11 2.1 Study subjects and data collection12 Seventy children (age = 2.5 (0.9-5.7) years old (median and range)), who were prescribed 3 months13 of fluticasone propionate treatment --based on Finnish guidelines for obstructive bronchitis-- were14 enrolled in our study at Tampere University Hospital. Each patient had IP and electrocardiography15 (ECG) signals recorded at home for three nights. The first recording (Week -1) was performed 1 week16 before conclusion of the fluticasone treatment, and the second (Week 2) and third (Week 4)17 recordings were performed 2 and 4 weeks after treatment ended. Recordings were obtained using a18 custom-designed device developed at Tampere University of Technology (Seppä et al., 2013b).19 Electrode placement was as previously described (Seppä et al., 2013a). On the first recording day20 (Week -1), a trained nurse placed the electrodes and the device on the patient at the hospital and21 instructed the parents on the procedure. For the following recordings (Week 2 and Week 4), the22 parents placed and activated the device at home. A nurse contacted the families to confirm the23 planned recording nights. In all cases, the device started recording before the patient went to sleep24 and recorded until after the patient woke up the next morning. On each recording day, parents25 photographed the electrode locations and noted the time of sleep onset, wake-up time, periods of26 nocturnal awakening, times of bronchodilator intake, and respiratory symptoms, usually coughing,27 sneezing, and rhinorrhoea. Patients were classified according to the following two classification28 criteria. For the first classification criteria, a paediatric pulmonologist followed the patients for 629 months after the last recording and classified them as current asthma (CA-Y) if they had been30 prescribed a regular asthma controller, reported difficult nocturnal coughing, exercise-induced31 coughing, or shortness of breath relieved by the bronchodilator; possible current asthma (CA-P) if they32 did not fulfil the preceding criteria but were prescribed intermittent controller medication for treating33 asthma symptoms; and no current asthma (CA-N) otherwise. For the second classification criterion,1 patients were classified as atopic if they responded positively in a skin-prick test against egg, cat, dog,2 birch, or timothy, or nonatopic otherwise. Classification criterion, demographic data, and3 bronchodilator use are summarised in Table 1. The Regional Ethics Committee of Tampere University4 Hospital approved the research protocol (Ethics Committee Code R14027), and the ethical guidelines5 of the Declaration of Helsinki were followed.6 2.2 Data preprocessing7 All the recordings were visually inspected by trained researchers who were blind to patient8 information. The researchers discarded sections corrupted by motion or other distorting events such9 as coughing, moving, or talking. Accepted sections of the recordings were automatically processed to10 derive minute-by-minute TBFV profiles, as previously detailed (Gracia-Tabuenca et al., 2019). In short,11 the ECG signal was used to filter out the cardiac artefact from the raw IP signal (Seppä et al., 2011). A12 Savitzky-Golay filter differentiated the resulting lung volume-oriented IP signal into a flow-oriented13 IP signal (Seppä et al., 2010), the remaining noise of which was further attenuated using a nonlinear14 projection filter (Gracia et al., 2017). Cleaned-up flow and volume IP signals were split into respiratory15 cycles, as recommended by Schmidt et al. (1998), and cycles were transformed into TBFV profiles.16 Resulting TBFV profiles were averaged in the flow-volume domain, as described by Sato and Robbins17 (2001), using a 20-TBFV moving window with a 5-TBFV overlap. Each averaged TBFV profile was18 normalised to unit volume and flow-scaled, making its time integral equal to 1 (Sato and Robbins,19 2001).20 For each profile, the following expiratory indices were measured as recommend by Bates et al., 2000;21 and Beydon et al., 2007: expiratory time (TE), time to peak tidal expiratory flow (TPTEF), their ratio22 (TPTEF/TE), equivalent volume ratio (VPTEF/VE), tidal expiratory flow when 50%, 25%, and 5% of the tidal23 volume remains in the lungs relative to peak tidal expiratory flow (TEF50/PTEF, TEF25/PTEF, and24 TEF05/PTEF, respectively). In addition, the index PFV was calculated as the exponent of a power25 function fitted between PTEF and TEF05, as described previously (Gracia-Tabuenca et al., 2019). Figure26 1 shows the indices measured in four representative profiles.27 2.3 Sleep segmentation28 The overnight recordings were segmented on the basis of two different methods: regions of high and29 low respiratory rate variability and time from sleep onset. Regions of high and low variability were30 automatically defined using a method similar to that proposed by Isler et al. (2016). In short, a31 respiration variability time series was formed using the median absolute deviation (MAD) of the32 interbreath intervals (IBI) within a moving 5-min 50% overlap window. Subsequently, a line was fitted1 to the variability time series and crossing points were marked. Regions of 5 min around the crossing2 points were discarded. The remaining sections with the most samples over the fitted line were defined3 as REM and NREM otherwise. An example of the process is shown in Figure 2. Although this4 implementation could not be validated against polysomnography, the performance of our method5 was putatively similar to that of the Isler et al. method (see the Discussion section). The time from6 sleep onset regions were defined as 3-h bins centred at each hour starting from sleep onset. A7 representative recording is shown in Figure 2. Sleep-onset time was set automatically as the beginning8 of the first segment that had no motion artefacts for more than 5 min. Only one automatic sleep onset9 was detected more than an hour before the time annotated by the parents. It was considered an error10 and the annotated time was used.11 2.4 Statistical analysis12 In each recording, we calculated for each index the median of all-night values within the REM sections13 and the median of all-night values within the NREM sections. Similarly, in each recording, we14 calculated for each index and for each hourly bin the median of the REM values and the median of the15 NREM values within each 3-h bin. If the number of indices within a bin was fewer than 20, that bin16 was rejected. The same two procedures were followed for IBI, heart rate, and MAD(IBI) signals. The17 following tests were performed on both: all-night medians (Table 2) and hourly bin medians (Figure18 3). Wilcoxon rank sum was used to assess the differences between the REM and NREM medians within19 each recording, separately for the three recording weeks. The same test was used to assess the20 differences between recording weeks for each subject, separately for REM and NREM. The differences21 between groups within each classification criterion and between bronchodilator use and no use were22 assessed for each index, in each sleep stage, and in each recording day using the Wilcoxon rank sum23 test. The characteristics of the subjects between groups or bronchodilator use were compared using24 the Kruskal-Wallis test for continuous variables or the c2/Fisher’s exact test for categorical variables25 (Table 1). Bonferroni correction was applied in all tests. Moreover, Spearman rank correlation26 coefficients were calculated between all-night median values and patient age for all weeks (Table 3).27 3. RESULTS28 Twenty-two recordings were rejected due to battery or electrode problems, or malfunctioning of the29 prototype recorders. For the accepted recordings, the starting time was at a mean of 9:20 pm (±1:05)30 and lasted 9.95 (±1.11) h (mean value (standard deviation). Of the accepted data, 28.33% (±4.47%)31 was discarded for being corrupted or occurring between sleep stages, and 25.17% (±5.93%) was32 classified as REM, slightly higher than reported by Traeger et al. (2005). Table 2 shows that neither1 faulty recordings nor sleep efficiency depended on the recording week, and summarises all-night2 medians for each index and recording week. Hourly bin medians for six selected indices and three3 recording weeks are summarised in Figure 3.4 MAD (IBI), heart rate, and respiratory rate showed similar results in on-medication and off-medication5 recording weeks. Evidently, IBI variability was higher during REM than during NREM the whole night.6 Only respiratory variability, respiratory rate, and TE presented a weak but significant correlation (p <7 0.05) with the patients’ age for some combinations of sleep stage and recording day. However, heart8 rate had a significant correlation with age (p < 0.001) for all stages and ages (Table 3). The correlation9 of age with heart rate, and less significantly with respiratory rate, agrees with published results10 (Scholle et al., 2011).11 For the on-medication recordings (Week -1), absolute times TE and TPTEF were both significantly shorter12 during REM for the whole night. However, their overnight median trends were different. The TE 13 median slightly increased overnight for both sleep stages, whereas the TPTEF median was constant for14 NREM and decreased for REM in the first part of the night. As expected, the overnight median trend15 for TPTEF/TE was the combination of the trend of TPTEF and the inverted trend of TE. For VPTEF/VE, the16 trend of TE was no longer present, but both NREM and REM presented a trend similar to that of TPTEF.17 However, unlike with TPTEF, for VPTEF/VE, the NREM and REM median trends overlapped with each other18 and, therefore, showed no significant differences overnight. On the end side of the TBFV profiles,19 TEF05/PTEF was significantly lower for REM than for NREM during the whole night. In the middle part20 of the profiles, PVF showed a constant median during the whole night for NREM and REM. On the other21 hand, TEF50/PTEF and TEF25/PTEF showed a decreasing trend, similar to that of VPTEF/VE, and a sleep-22 stage differentiation similar to that of TEF05/PTEF. The time progression of TEF50/PTEF and23 TEF25/PTEF seemed to be a combination of VPTEF/VE and TEF05/PTEF (not shown in Figure 3).24 A comparison of the indices for on-medication (Week -1) with those for off-medication (Week 2 and25 Week 4) showed there were no significant differences in IBI variability, heart rate, and respiratory26 rate. Overnight median trends for TE, TPTEF, TPTEF/TE, and VPTEF/VE presented night progressions for off-27 medication similar to those for on-medication. However, in the Week 4 recordings, all-night medians28 for TPTEF, TPTEF/TE, and VPTEF/VE were significantly lower during NREM. The most significant difference29 was observed for VPTEF/VE (p = 0.0019), which also showed a significant decrease for all hourly bin30 medians. Likewise, TEF05/PTEF increased in the Week 4 recordings for both sleep stages, but the31 increase was statistically significant for all hourly bin medians only for NREM. Unlike in Week -1, PVF 32 significantly increased in Week 4 for both sleep stages, but only in the second half of the night.1 Moreover, the increase in PVF was higher for REM than for NREM.2 Current asthma and skin-prick test classifications showed no significant differences for any index in3 any sleep stage on any recording day. However, the use of a bronchodilator showed significant4 differences (p < 0.01) in Week 4 for both sleep stages for all the indices in the middle part except5 TEF25/PTEF in NREM. Counterintuitively, the values of subjects who used a bronchodilator suggest6 that they presented greater obstruction than subject who did not use: during REM, PVF was 0.89 (0.837 1.04) (median (interquartile range)) for bronchodilator use vs. no use 0.71 (0.67 0.77); TEF25/PTEF8 was 0.44 (0.39 0.46) vs. 0.53 (0.50 0.57); and TEF50/PTEF was 0.74 (0.68 0.77) vs. 0.82 (0.79 0.87).9 Similarly, during NREM, PVF was 0.82 (0.73 0.92) vs 0.69 (0.64 0.74) and TEF50/PTEF was 0.77 (0.7510 0.81) vs. 0.85 (0.80 0.89).11 12 4. DISCUSSION13 This study demonstrated that dividing the night into regions of higher and lower IBI variability, as an14 estimation of REM and NREM sleep, presented differences in the TBFV indices for both on-medication15 and off-medication recordings in children. Moreover, when assessed at different times from sleep16 onset, certain indices presented a decreasing averaged trend during REM. In addition, the interruption17 of treatment had a different effect on the early and late parts of the expiratory TBFV profile than on18 the middle part. Changes in the early and late expiration were significant during the whole night for19 NREM. Changes in the middle expiration were significant in the second part of the night and larger for20 REM.21 Lower TE and TPTEF values for REM than for NREM have been observed in healthy and wheezing infants22 (Gracia-Tabuenca et al., 2019; Haddad et al., 1979), but not in adolescents (Tabachnik et al., 1981).23 We found that the ratio TPTEF/TE was lower for REM than for NREM but that VPTEF/VE was similar for24 both sleep stages. Such different results for these similar ratios can be explained by comparing the25 late part of expiration on the time and volume domains. For example, the profiles in Figure 1 (A) and26 (B) present similar VPTEF/VE, but TPTEF/TE is lower in (B) because in the late part, expiratory airflow is27 low. Hence, a longer time is needed to produce the same change in volume as in (A), where the flow28 is higher. Thus, our results suggest that for REM sleep, the later part of exhalation was interrupted29 less often, whereas for NREM, exhalation was interrupted more often before reaching resting volume,30 as is also suggested by a higher TEF05/PTEF during REM. Shorter and uninterrupted exhalation during31 REM may be due to the natural decrease in respiratory musculature tone in this sleep stage (Horner32 R.L., 2010). Intercostal atony in REM leads to a more compliant chest that deflates faster (Mortola et1 al., 1982; Otis et al., 1950). Intercostal atony together with a lower diaphragm tone decreases the2 functional residual capacity (FRC) (Henderson-Smart and Read, 1979). This decrease has been linked3 to uninterrupted or late interruption of expiration (Morris et al., 1998; Schmalisch et al., 2003).4 The overnight decreasing trend in the REM bin medians, which is shared by TPTEF, TPTEF/TE, VPTEF/VE,5 TEF50/PTEF, and TEF25/PTEF, may have been caused by a shortening of post-inspiration inspiratory6 activity (PIIA) during the night. For individuals of all ages, a decrease in TPTEF, and therefore in TPTEF/TE 7 and VPTEF/VE, is commonly understood as a shortening of PIIA (Ent et al., 1998). Shorter PIIA would also8 explain the lower TEF50/PTEF and TEF25/PTEF values because decreased expiratory braking leads to9 higher PTEF (Walraven et al., 2003). The shortening of PIIA during the night may be due to multiple10 factors such as a decrease in respiratory musculature tone during the night, as seen in asthmatic adults11 (Steier et al., 2011); an adaptation to a circadian increase in airway resistance (Bellia et al., 1989); or12 other circadian factors (Bohadana et al., 2002). In any case, changes in the sleep stage or night13 progression did not seem to affect the number of concave profiles, as assessed by PVF, or the14 interruption of expiration, as assessed by TEF05/PTEF, at least for Week -1.15 Changes in the off-medication TBFV profiles compared to the on-medication TBFV profiles were16 presumably caused by an increased number of children presenting airflow limitation. Such changes in17 the early, middle, and late parts of expiration agreed with the changes related to airflow limitation18 reported in the following studies. In the early part, the significant decrease in VPTEF/VE, TPTEF/TE, and19 TPTEF was potentially caused by a shortening of PIIA. It has been speculated that individuals with airway20 obstruction have short PIIA braking to accommodate for the slower passive expiration (Carlsen and21 Carlsen, 1994; van der Ent et al., 1996). In the late part, the significant increase in TEF05/PTEF may be22 due to the early interruption of expiration with the purpose of elevating the FRC to increase airway23 calibre (Greenough et al., 1989; Wheatley et al., 1990). In the middle part, the significant increase in24 PVF was most likely due to an increase in concavity, as observed in infants (Benoist et al., 1994) and25 adults (Williams et al., 1998). Bronchodilator use decreases airway obstruction, thus putatively making26 the TBFV profiles less concave. However, our results show that profiles were more concave the days27 where bronchodilator was used. This apparent contradiction can be explained as follows.28 Bronchodilator use occurred mostly before the recording period and its effects are known to wear off29 after a few hours. Therefore, any changes in the profiles because of bronchodilator use were likely30 averaged out over the rest of the recording. Under these assumptions, bronchodilator use indicates31 that on that recording day, parents notice airflow limitation and applied the medication, but for most32 of the recording bronchodilator had no effects. This, together with the lack of correspondence33 FIGURE LEGENDS1 2 3 Figure 1. TBFV indices for four representative expiration limbs from the same patient extracted from4 NREM Week -1 (A), REM Week -1 (B), NREM Week 4 (C), and REM Week 4 (D). Upper plots show flow-5 time domain and lower plots show flow-volume domain . Volume is normalized to 1 and flow is scaled6 to have area of 1 in the flow-time domain. Light grey lines represent the expiration signals, vertical7 solid lines between points show distances, and solid line curves are the power-fitted curves, which are8 displaced for clarity and the dotted lines project where the curves should be located. The grey areas9 in the flow-time plots (A) and (B) are regions with the same area. They show that integrating the same10 volume (. ) took a longer time in (B) than in (A) because the flow () was lower (=∫d).11 1 Figure 2. Two sleep segmentation methods: (upper) segments of high and low respiratory rate2 variability are shown as grey and light-grey boxes, respectively; (lower) solid lines indicate time from3 sleep onset in 3-h segments. The procedure presented in the text is based on the dotted-line signal4 (MAD(IBI)). The dashed line is the linear fit and the black dots are valid crossings. The abscissa shows5 the time from sleep onset in hours.6 7 1Figure 3. Averaged hourly progression of several indices grouped by recording week and sleep stage.2 Rows correspond to an index and the columns to a recording week. Within each plot, the x-axis is the3 time from sleep onset (in hours) and the y-axis is the index value. Dots and vertical lines are median4 and interquartile ranges of all patients at a given time for NREM (grey) and REM (black). *: significant5 difference (p < 0.01) between sleep stages; a: significant difference (p < 0.05) between Week -1 and6 Week 4; b: significant difference (p < 0.05) between Week 2 and Week 4. Letters on top of vertical7 lines for NREM and letters on bottom for REM. All p values were calculated using the Wilcoxon signed8 sum test after Bonferroni corrections (n = 3).9 Table 1. Characteristics of studied children.1 Current a s t h ma Skin prick Groups CA - N CA - P CA - Y N on atopic A topic Subjects 16 16 36 40 28 Age [mo] 51.45 (18.23 77.07) 54.27 (40.07 80.70) 41.10 (15.73 78.63) 43.87 (17.77 67.50) 51.23 (15.73 80.70) Male 12 10 22 26 18 Broncho - Week - 1 2 0 2 2 2 Broncho - Week 2 1 4 7 9 3 Broncho - Week 4 3 6 7 11 5 2 Subjects were classified according to two classification criteria: current asthma and skin-prick test.3 The first criteria consist on tree groups: no current asthma (CA-N), probable current asthma (CA-P),4 and current asthma (CA-Y). The second criteria consist on two groups: nonatopic and atopic. Age (in5 months) is given as the median (range). The entries for Broncho-Week -1, Broncho-Week 2, and6 Broncho-Week 4 are the number of subjects who used a bronchodilator in Week -1, Week 2, and Week7 4, respectively. No significant difference was found between groups within each criterion for any8 characteristic as determined by the Kruskal-Wallis test (continuous variables) or the c2/Fisher’s exact9 test (categorical variables).10 Table 2. Median of the TBFV parameters during estimated NREM and REM sections overnight. Values1 are grouped according to the recording week.2 Week -1 Week 2 Week 4 Subjects 62 64 62 T REM /T Total [%] 0.25 (0.22 0.29) 0.24 (0.21 0.28) 0.24 (0.21 0.29) MAD(IBI) [s] NREM 0.13 (0.12 0.14) 0.13 (0.11 0.15) 0.13 (0.12 0.14) REM 0.28 (0.25 0.34) * 0.30 (0.26 0.33) * 0.29 (0.25 0.34) * Heart rate [bpm] NREM 83.30 (77.37 89.86) 83.53 (75.47 92.11) 84.94 (79.66 92.95) REM 89.90 (82.54 95.80) * 89.07 (81.85 97.04) * 89.41 (84.91 97.91) * Mean(IBI) [bpm] NREM 19.33 (18.32 22.35) 19.57 (17.89 21.85) 20.47 (18.21 23.04) REM 20.44 (18.17 22.01) 19.92 (18.40 22.69) 20.93 (18.93 23.62) TE [s] NREM 1.85 (1.65 2.00) 1.87 (1.71 2.04) 1.82 (1.59 2.00) REM 1.71 (1.59 1.92) * 1.75 (1.59 1.96) * 1.64 (1.50 1.94) * TPTEF [s] NREM 0.32 (0.27 0.37) 0.32 (0.27 0.37) 0.29 (0.24 0.34) a REM 0.27 (0.24 0.32) * 0.27 (0.24 0.29) * 0.26 (0.22 0.29) * TPTEF/TENREM 0.17 (0.15 0.20) 0.17 (0.14 0.20) 0.16 (0.14 0.19) a REM 0.16 (0.14 0.19) * 0.15 (0.13 0.18) * 0.15 (0.13 0.18) * VPTEF/VENREM 0.25 (0.22 0.29) 0.24 (0.21 0.28) 0.23 (0.20 0.27) a REM 0.26 (0.23 0.28) 0.24 (0.21 0.26) 0.23 (0.21 0.27) a TEF50/PTEF NREM 0.87 (0.84 0.89) 0.85 (0.80 0.88) 0.84 (0.79 0.88) a, Br REM 0.85 (0.81 0.88) * 0.81 (0.79 0.86) * 0.80 (0.76 0.87) *, a, b, Br TEF25/PTEF NREM 0.57 (0.53 0.61) 0.57 (0.51 0.60) 0.53 (0.48 0.60) REM 0.54 (0.52 0.57) * 0.52 (0.48 0.56) * 0.50 (0.45 0.56) *, a, b, Br PVF NREM 0.68 (0.64 0.73) 0.69 (0.65 0.75) 0.71 (0.66 0.80) a, Br REM 0.69 (0.65 0.75) 0.72 (0.67 0.78) 0.75 (0.68 0.85) *, a, b, Br TEF05/PTEF NREM 0.13 (0.09 0.16) 0.14 (0.10 0.18) 0.14 (0.10 0.21) a REM 0.11 (0.09 0.14) * 0.11 (0.08 0.13) * 0.12 (0.08 0.16) *, a Values are given as median (0.25 0.75 (quartiles)). Columns are recording weeks: Week -1 is one week3 before end of treatment, Week 2 is two weeks after end of treatment, and Week 4 is 4 weeks after4 end of treatment. Indices are defined in the text. *: significant difference (p < 0.01) between sleep5 stages within each week. a: significant difference (p < 0.05) between Week -1 and Week 4. b:6 significant difference (p < 0.05) between Week 2 and Week 4. Significant differences were calculated7 using the Wilcoxon signed sum test. Br: significant difference (p < 0.01) between subjects who used a8 bronchodilator that recording day and subjects who did not, calculated using the Wilcoxon rank sum9 test. Bonferroni correction (n = 3) was applied to all p values.10 11 Table 3. Spearman correlation coefficients between selected indices and patient age during NREM and1 REM recorded in the week under treatment (Week -1).2 Week -1 Week 2 Week 4 MAD(IBI) [s] NREM 0.29 (0.02) * 0.35 (0.01) * 0.13 (0.33) REM -0.01 (0.97) 0.10 (0.45) -0.01 (0.96) Heart rate [bpm] NREM -0.42 (0.00) ‡-0,40 (0.00) ‡-0.49 (0.00) ‡ REM -0.42 (0.00) ‡-0.44 (0.00) ‡-0.50 (0.00) ‡ Respiration rate [bpm] NREM 0.17 (0.17) 0.13 (0.03) * 0.20 (0.13) REM 0.26 (0.04) * 0.23 (0.04) * 0.33 (0.01) * TE [s] NREM 0.17 (0.19) 0.13 (0.30) 0.19 (0.14) REM 0.21 (0.09) 0.23 (0.07) 0.34 (0.01) * Correlation tested using Spearman’s rank correlation rho (r). *: p< 0.0001, ‡: p < 0.0001. The indices3 not included in the table had a nonsignificant correlation with p > 0.05.4