Polysomnography tracings with neurally adjusted ventilatory assist (NAVA) and pressure support ventilation (PSV) in a representative subject. C4-A1 and O3-A2 are the electroencephalogram readings. ROC and LOC are the electrooculogram readings. Chin is the electromyogram reading. P peak is the peak airway 

Context in source publication

Context 1

... a prospective, interventional, crossover study in a 22-bed medical ICU at H ˆpital du Sacr ́-Cœur over a 12-month period. Inclusion criteria were invasive mechanical ventilation, normal consciousness, absence of sedatives and opiates for Ն 24 h, and PSV with an F IO 2 below 0.60 and a PEEP of 5 cm H 2 O to achieve an S pO 2 of at least 90%. Exclusion criteria were presence of a central nervous system disorder, Glasgow Coma Scale score Ͻ 11, hemo- dynamic instability, renal and/or hepatic insufficiency, on- going sepsis, history of esophageal varices or gastroesoph- ageal bleeding in the past 30 days, and history of gastrointestinal symptoms such as vomiting. All subjects were ventilated through an endotracheal or tracheostomy tube. Subjects meeting the inclusion criteria were connected to a ventilator capable of delivering both PSV and NAVA (Servo-i, Maquet Critical Care, S ̈lna, Sweden). As previously described, 19 EAdi was recorded using a 16 French gastric catheter equipped with electrodes (EAdi catheter, Maquet Critical Care, S ̈lna, Sweden), whose correct position was checked using the “EAdi catheter positioning” ventilator function. Humidification was provided by a heat and moisture exchanger in all subjects. End-tidal CO 2 (P ETCO 2 ) was monitored with the Servo-i volumetric CO 2 module. PSV and NAVA were delivered in random order, determined using a closed-envelope technique. Each mode was delivered for 4 periods of 4 hours: 2 during the day (7:00 to 11:00 AM and 12:00 to 4:00 PM ) and 2 at night (10:00 PM to 2:00 AM and 3:00 to 7:00 AM ). To prevent possible data contamination from the previous mode of ventilation, data acquisition was delayed for 1 hour after each ventilator setting change (Fig. 1). Inspiratory pressure support was titrated to obtain a V T of 8 mL/kg of predicted body weight during active inspi- ration, without exceeding 20 cm H 2 O of pressure support and a breathing frequency of Ͻ 35 breaths/min. Flow- trigger sensitivity was set as low as possible (1–2 L/min) without inducing auto-triggering; cycling-off occurred when the inspiratory flow fell below 25% of the peak inspiratory flow. During PSV the ventilator function “NAVA Preview” was used to estimate the NAVA level required to achieve the same peak inspiratory pressure. After switching to NAVA, the settings were gradually adjusted to deliver the same assist level (peak pressure) and external PEEP as used previously with PSV. With NAVA the EAdi inspiratory trigger was at the lowest setting (0.5 ␮ V). EAdi was recorded with both ventilation modes and used to evaluate patient-ventilator synchrony. The backup apnea ventilation delay was set at 15 seconds, so that apneas lasting 10 seconds or more could be identified and their impact on sleep architecture (arousals) assessed. The electroencephalogram was recorded using standard electrode positions (left frontal/right auricular reference [F3/A2], right frontal/left auricular reference [F4-A1], left central/right auricular reference [C3/A2], right central/left auricular reference [C4/A1], left occipital/right auricular reference [O1/A2], and right occipital/left auricular reference [O2/A1] in the international 10 –20 system for electrode placement). 20 The standard reference was the left mastoid lead. 20 Two electrooculogram and 3 chin electromyogram leads were used to score rapid-eye-movement (REM) and non-REM sleep. The electroencephalogram, right and left electrooculograms, and submental electromyogram signals were amplified and recorded in the data acquisition system (Alice 5 polysomnography system using Alice Sleepware 2.5 software, Respironics, Murrys- ville, Pennsylvania). Sleep recordings were scored manually by a pulmo- nologist who was blinded to the study data and used Rechtschaffen and Kales criteria for sleep stages, 21 and American Sleep Disorder Association criteria for arousals and awakenings. 22,23 Central apneas were diagnosed ac- cording to international recommendations, 23 as absence of breathing and respiratory effort for at least 10 seconds (Fig. 2). Arousals and awakening were classified as apnea- related if they occurred within 3 breaths and/or 15 seconds of the end of an apnea. 24,25 Flow and pressure signals were acquired at a sampling rate of 100 Hz. Signals from each electrode pair were differentially amplified, digitized, and processed online, using previously described filters and algorithms. 26-28 Changes in diaphragm position along the array were accounted for using the cross-correlation technique, 29,30 and diaphragm-to-electrode distance filtering was minimized using the double-subtraction technique. 31 The root mean square was calculated for the subtracted signal and the signal obtained from the electrode pair on the diaphragm, and these values were summed every 16 ms to quantify EAdi. Signal segments with residual disturbances due to cardiac electric activity or common mode signals were identified using specific detectors and replaced by the previously accepted value. This processed EAdi signal was used to control the ventilator during NAVA and was si- multaneously acquired at a sampling rate of 2,000 Hz. EAdi was obtained from the ventilator through an RS232 interface, at a sampling rate of 100 Hz, and recorded using dedicated software (Nava Tracker 2.0, Maquet Critical Care, S ̈lna, Sweden). The input was analyzed using software (Analysis 1.0, Maquet Critical Care, S ̈lna, Sweden, and a customized version of Excel, Microsoft, Redmond, Washington). The peak of rectified and integrated EAdi swings (peak EAdi) was measured. From the flow signal we obtained the ventilator frequency (f-flow), flow-based inspiratory time, and flow-based expiratory time. To evaluate breathing pattern variability during the different sleep stages with each mode, we calculated the coefficient of variability (defined as the standard error/mean ratio ϫ 100) for V , breathing frequency, and peak EAdi. 32,33 The statistical analysis was performed using statistical software (SPSS 17.0, SPSS, Chicago, Illinois). Continuous variables are described as mean Ϯ SD. Data were compared using the general linear model for repeated measures. Given the small sample size, a nonparametric test was used to compare variables; we chose the Wilcoxon test for paired samples. Two-tailed P values smaller than .05 were considered significant. Variability data were compared using analysis of variance for repeated measures. We included 14 subjects, whose main characteristics at admission are reported in Table 1. Mechanical ventilation was required for acute respiratory failure in 10 subjects, postoperative cardiac complications in 3 subjects, and sep- tic shock in 1 subject. Mean PSV level was 15 Ϯ 5 cm H 2 O, and mean NAVA level was 1.6 Ϯ 1.4 cm H 2 O/ ␮ V. PEEP was kept at 5 cm H 2 O in all subjects with both ventilation modes. Mean total sleep time was 537 193 min. Mean sleep efficiency (the percentage of sleep during the study) was 56 20%. Table 2 reports the main breathing pattern data. Mean V ̇ E did not differ significantly between PSV and NAVA: 9.6 Ϯ 1.8 L/min and 9.3 Ϯ 1.8 L/min, respectively ( P ϭ .51). Mean breathing frequency was 17 Ϯ 4.6 breaths/min with PSV, and 18 Ϯ 5.7 breaths/min with NAVA ( P ϭ .14). The mean V ̇ E , V T , and breathing frequency did not differ significantly between the 2 ventilation modes. P ETCO 2 was not significantly different between the 2 modes during any of the sleep stages. V T differed significantly between the 2 ventilation modes during sleep stage 3– 4 (mean 391 Ϯ 57 mL with NAVA and 435 Ϯ 64 mL with PSV, P ϭ .005) and during REM sleep (mean 360 Ϯ 54 mL with NAVA and 415 Ϯ 61 mL with PSV, P ϭ .008, see Table 2). During PSV, V T did not vary significantly between wakefulness (429 Ϯ 65 mL) and REM sleep (415 Ϯ 61 mL) ( P ϭ .08). However, with NAVA, V T was significantly greater during wakefulness (419 Ϯ 63 mL) than during REM sleep (360 Ϯ 54 mL) ( P ϭ .001). V T differed significantly between stage 1 and REM sleep with both modes (NAVA 431 Ϯ 69 mL vs 360 Ϯ 54 mL, respectively, P ϭ .001, PSV 442 Ϯ 61 mL vs 415 Ϯ 61 mL, respectively, P ϭ .001, see Table 2). Mean breathing frequency during non-REM sleep did not differ significantly between the 2 ventilation modes. During REM sleep, in contrast, mean breathing frequency was significantly higher with NAVA than with PSV (15 Ϯ 3 breaths/min and 13 Ϯ 2 breaths/min, respectively, P ϭ .004). During wakefulness the mean breathing frequency was also significantly higher with NAVA (20 Ϯ 4 breaths/min and 18 Ϯ 3 breaths/min, respectively, P ϭ .005, see Table 2). Table 3 shows the oscillatory behavior of V T , breathing frequency, V ̇ E , and P ETCO 2 during sleep stages 2 and 3– 4 with PSV. Sleep apnea occurred in 10 subjects, all of whom had either COPD or chronic heart failure, 2 known risk factors for central apneas. Moreover, these subjects experienced hyperventilation during sleep, which preceded the central apneas. Table 3 reports the mean number of sleep apneas per hour of sleep in the ...