Patients admitted over a period of six months to the ICUs of the University of Bari and Ferrara Academic Hospitals were considered for enrollment in the study. The study was approved by the Ethics Committee of the Azienda Ospedaliero-Universitaria Policlinico di Bari (protocol no. 257) and of the Arcispedale Sant’Anna hospital, Ferrara, Italy (protocol no. 131084). Informed consent was obtained from each patient according to local regulations. The study was conducted between January 2016 and July 2016 in accordance with the Declaration of Helsinki. A physician not involved in the study was always present for patient care.
Patients were eligible for the study if they were older than 18 years and excluded if they were affected by neurological or neuromuscular pathologies, had known phrenic nerve dysfunction or any contraindication to the insertion of a naso-gastric catheter (for example: recent upper gastrointestinal surgery, esophageal varices).
Patients were studied in the semi-recumbent position. All the patients were ventilated with a Servo i ventilator (Maquet Critical Care, Solna, Sweden) equipped with the neurally adjusted ventilatory assist (NAVA) software that includes the “neuro-ventilatory tool” for EAdi measurement. At the beginning of the study, the standard naso-gastric tube was replaced with a 16 Fr, 125 cm, EAdi catheter (Maquet Critical Care, Solna, Sweden) unless an EAdi catheter was already in place. The EAdi catheter was first positioned according to the corrected nose-ear lobe-xyphoid distance formula and subsequently through the EAdi catheter position tool (Servo i, NAVA software) .
Peak airway opening pressure (PAO PEAK) and positive end-expiratory pressure (PEEP) were measured from the PAO signal. Tidal volume (VT) was measured as the integral of the inspiratory flow. Mechanical respiratory rate (RR), inspiratory and expiratory time (Ti,MECH and Te,MECH, respectively) were measured by the flow and PAO signals. The inspiratory EAdi peak (EAdiPEAK), the slope of the EAdi from the beginning of inspiration to the peak (EAdiSLOPE) and the neural inspiratory time (TiNEUR) were measured from the EAdi waveform.
The inspiratory pressure generated by the diaphragm (trans-diaphragmatic pressure, PDI) was calculated according to the method recently validated by Bellani and coworkers [7,18,21,31]. Briefly, we first calculated the diaphragmatic neuro-muscular efficiency (NME) as the ratio between the negative deflection peak in PAO during a spontaneous inspiratory effort (recorded during a brief end-expiratory occlusion lasting 5-10 s) and the corresponding peak in the EAdi curve. The NME measures the diaphragmatic neuro-mechanical coupling, and can be used to convert the EAdi into PDI (PDI = EAdi * NME)  . The inspiratory PDI pressure-time product per breath (PTPDI/b) was calculated as the area under the PDI signal. The inspiratory PDI pressure-time product per minute (PTPDI/min) was calculated as:
PTPDI/min = PTPDI/b * RR.
The breathing pattern and EAdi- parameters, obtained from the RS232 port of the Servo i ventilator at a sampling rate of 100 Hz, were stored in a personal computer (NAVA tracker software, Maquet Critical Care, Solna, Sweden) for subsequent analysis (ICU Lab automatic analysis software, Kleistek Engineering; Bari, Italy).
Patients were admitted to the study within 24 hours after the shift in the PSV mode. At the beginning of the study, the PSV level was carefully titrated to obtain a VT between 5 and 8 ml/Kg predicted body weight (PBW) and a RR between 20 and 30 breaths/min [6,34,35]. The inspiratory trigger was set in the flow-by mode, sensitivity level 5 (Servo i arbitrary units); the expiratory trigger was set at 30 % of the peak inspiratory flow. Clinical PEEP and FiO2 levels were left unchanged. Starting from the end of the PSV titration phase, patients were studied for 12 hours, from 8 a.m. to 8 pm.
In order to continuously assess the neuro-ventilatory drive throughout the 12 hours study period, the EAdi waveforms were analyzed through the automatic EAdi analysis software, a dedicated function of the ICU Lab software (Kleistek Engineering; Bari, Italy). This software identifies the EAdi peaks corresponding to each single breath and transfers the EAdi related data (EAdiPEAK, EAdiSLOPE, TiNEUR) in an excel sheet. Since the breathing pattern parameters (VT, RR, PaoPEAK, TiMECH) could not be examined continuously by the software, for each patient we analyzed manually the first 30 consecutive breaths available for each neuro-ventilatory drive class through the dedicated function of the Kleistek software.
Based on previous studies [7,25,28,29] and on the manufacturer instructions (Maquet Critical Care AB, NAVA flow chart MX-6462 Rev 02/2015), we pre-defined three neuro-ventilatory drive classes: “Low”, for breaths with EAdiPEAK below 5 mV; “Normal”, for breaths with EAdiPEAK between 5 and 15 mV and “High” for breaths with EAdiPEAK higher than 15 mV.
Patient-ventilator asynchronies were assessed by taking into account in the first 20 consecutive min, for each EAdi class, based on the method proposed by Thille and coworkers . Asynchronies were classified into six types: a) ineffective triggering (missed effort); b) ineffective inspiratory triggering; c) double-triggering; d) auto-triggering; e) prolonged cycle; f) short cycle . The Asynchrony Index (AI) was calculated as:
AI = Total number of asynchronies/(mechanical cycles + missed efforts).
We assessed the number the percentage of time spent in each of the three pre-defined EAdiPEAK classes (i.e. “Low”, “Normal” and “High”). Differences between percentages were analyzed through the chi-square test. In order to estimate the average time spent in each of the three EAdi classes, we applied the Generalized Estimated Equation (GEE) model . In the GEE model the single breath is the first level unit, the time of each breath is the dependent variable, the class of EAdi is the independent variable and, finally, the patient is the second level unit. Pairwise comparisons between the estimate times spent in each of the three neuro-ventilatory drive classes were adjusted according to Tukey.
Normally distributed continuous data are expressed as means and standard deviation (SD) and non-normally distributed data are expressed as median and interquartile range (IQR). Normality of continuous data was tested through the Kolmogorov–Smirnov test. The ANOVA or the Friedman repeated measure analysis of variance was used as appropriate. Pairwise comparisons were adjusted according to Tukey.
A multivariable multinomial logistic model for ordinal variables and repeated measures was applied to evaluate the effect of TiMECH, PaoPEAK, VT/PBW, RR and PS level on the probability of being in one of the three EAdiPEAK classes. All the statistical tests were two-tailed, and p-values of less than 0.05 were considered statistically significant. Statistical analysis was performed by software SAS 9.4 (SAS Institute, Cary NC).