METHODS

Patients

From December 2008 to April 2009, we studied 10 ventilator-dependent patients with neuromuscular disease during their usual respiratory follow-up in the R. Poincaré teaching hospital (Garches, France). All patients received mechanical ventilation via a cuffless tracheostomy. Pulmonary function tests routinely performed for the follow-up were collected; all measurements were made with a mouthpiece after occlusion of the tracheostomy tube.

The study protocol was approved by our institutional review board, and written informed consent was obtained from all patients before study inclusion.

Experimental setup

Ventilator-delivered flow was measured using a pneumotachograph (Fleisch #2, Lausanne, Switzerland) and tracheal pressure at the proximal end of the tracheostomy tube using a differential pressure transducer (Validyne MP 45±100 cm H2O, Northridge, CA, USA). To assess patient gas exchange, oxygen saturation (SpO2) was estimated using pulse oximetry (Ohmeda Biox, BOC Healthcare, Boulder, CO, USA).

Acoustic speech signals were recorded using three methods. The signals recorded from a microphone (DM202, MDE, Pierron, Sarreguemines, France) positioned 20 cm from the patient’s lips were routed to a microcomputer with an AD converter (MP150® and Acqknowledge®, Biopac system, Goleta, CA, USA) that synchronized respiratory data (ventilator flow and tracheal pressure) and acoustic data. The AD converter digitized respiratory signals at 128 Hz and speech signals at 20 000 Hz. The acoustic signal was also routed to the Dragon NaturallySpeaking 10® speech recognition system (Nuance; Burlington, MA, USA) with its specific microphone and a laptop computer containing its speech recognition software. Finally, the acoustic signal was recorded on a digital recorder (DS55, Olympus®, Herodphot, Manage, France) with a signal bandwidth between 50 and 19000 Hz to allow assessment of fundamental frequency and subsequent qualitative analysis by speech therapists.

Experimental protocol

All patients were usually ventilated in the ACV mode. VI, TI, backup rate, and inspiratory trigger sensitivity were kept unchanged. Before testing, patients were familiarized with the use of 5 cm H2O PEEP and of the PMV. These two conditions were tested in random order; patients were given 20 minutes to adjust to each techniques. With each condition, the patient first used the Dragon NaturallySpeaking 10® voice-training menu, which involves reading a text passage for 10 minutes to enable software training. Then, the patient continuously uttered the [a] sound for as long as possible, read a list of words, uttered the [a] sound in a glissando from high pitch to low pitch then from low pitch to high pitch, and read a standard text passage.

Data analysis

RR, TI, VI, and the volume expired through the tracheostomy tube (VE) were measured based on the computerized flow signal. The difference between VI and VE was used as an approximation of the volume expired through the upper airway. Ventilator triggering by the patient was considered significant when RR exceeded the backup rate by at least 3 cycles/min.

Speech was evaluated by measuring the mean time spent speaking during the respiratory cycle, time needed to read the text passage, ability of Dragon NaturallySpeaking 10® to accurately recognize the spoken words, and perceptual analyses by two speech therapists blinded to speech condition. The numbers of grammatical and phonological errors made by Dragon NaturallySpeaking 10® during the standard text reading were compared between PEEP and PMV. The two speech therapists had no direct experience working with ventilator-supported patients. They assessed the recordings presented in pairs, with PEEP vs. PMV in random order. The intelligibility scale used was a French adaptation of the Frenchay Dysarthria Assessment (0 to 8 scale) evaluated with an inter judge reliability of 92.5 % and in intra-judge reliability (at 1- point) of 87% [15]. In addition, the speech therapists determined the most used French perceptual score on a 0 to 128 scale, validated, with an inter judge reliability of 80 % and in intra-judge reliability (at 1- point) of 90 % [16].

At the end of each trial, the patients evaluated subjective respiratory comfort and subjective speech comfort on a 10-cm horizontal visual analog scale (VAS) [17, 18].

Statistical analysis

All results are expressed as median and interquartile range (25%-75%). As the samples were too small for an assessment of normality, we used non-parametric tests. Differences between the two conditions were assessed using the Wilcoxon test. P values <0.05 were considered statistically significant.