Mechanical ventilation is a prominent lifesaving treatment. It is, however, associated with an array of adverse effects, which include ventilator-associated pneumonias, volume-induced lung injury and, more recently identified, ventilator-induced diaphragm dysfunction (VIDD) [1-3]. VIDD combines diaphragm weakness with muscle fibre atrophy, remodelling and injury. Its mechanisms involve decreased protein synthesis, increased proteolysis, increased oxidative stress and mitochondrial dysfunction [2, 4]. Controlled mechanical ventilation appears to be the main, if not the sole, risk factor for VIDD, which in animal models is attenuated by the maintenance of respiratory efforts (assisted ventilatory modes) [2]. Although the corresponding human evidence is still lacking, this underlies the current notion that ''clinicians should encourage persistent diaphragmatic activity'' in patients receiving mechanical ventilation [2]. Diaphragm pacing has been proposed as a surrogate for spontaneous respiratory activity when the latter is not compatible with the condition of the patient [2, 5], but this approach has, seemingly, not yet been tested experimentally. Here we report a preliminary description of putative beneficial effects of diaphragm pacing in three mechanically ventilated sheep. Three female adult sheep (41, 32 and 34 kg), were anaesthetised (premedication: acepromazine 1.3 mg?kg-1 i.m., 30 min before induction; induction: propofol 6 mg?kg-1 i.v.; maintenance: continuous propofol 1-2 mg?kg-1 ?h-1 , midazolam 0.3-2 mg?kg-1 ?h-1 and morphine 0.2-0.3 mg?kg-1 ?h-1 ; no paralysing agents), tracheotomised and mechanically ventilated with a minute ventilation ensuring normocapnia (Aisys, GE Healthcare, Datex Ohmeda, Madison, WI, USA). Additional oxygen was given to maintain transcutaneous-pulsed oxygen saturation .92%. Adequate fluid and nutritional support was provided and glycaemia controlled. Body temperature, heart rate and arterial pressure were monitored. Intradiaphragmatic phrenic nerve stimulation electrodes were inserted bilaterally in the hemidiaphragms using the cervical incision thoraco-endoscopic approach (CITES) [6]. Diaphragm pacing (NeurRxDP4 stimulator, Synapse Biomedical Inc., Oberlin, OH, USA) was instituted within 2 h following the initiation of mechanical ventilation. Only one hemidiaphragm was stimulated throughout the experiments (right in two cases, left in one case), allowing comparison between mechanically ventilated hemidiaphragms (nonstimulated side) and mechanically ventilated with diaphragm pacing (MV+stim) ones (stimulated side). 30 min stimulation sessions were superimposed upon mechanical ventilation at 4 h intervals (18 breaths per minute, stimulation intensity 15 mA, stimulation frequency 20 Hz, inspiratory time 1.1 s and pulse width 150 ms, mechanical and electrical inspirations were synchronous). One animal suddenly died after 48 h of mechanical ventilation (#3) and the two others (#1 and #2) were sacrificed after 72 h of mechanical ventilation. Costal diaphragm strips measuring 1068 cm were taken through an abdominal incision, immediately post mortem in animal #3 and immediately before sacrifice in animals #1 and #2. Diaphragm samples were mounted on a small mound of 10% Gum Tragacanth placed on a cork disc and frozen in isopentane cooled with liquid nitrogen. Transverse serial cryosections (8 mm thick) of the costal diaphragm were stained with haematoxylin and eosin, and analysed qualitatively for structural abnormalities. Other sections were stained for adenosine triphosphatase (ATPase pH 9.4) and fibre types were identified, according to their histochemical reactions, as slow twitch type I or fast-twitch type II fibres using ImageJ software (US National Institutes of Health, Bethesda, MD, USA). For each fibre type, an average cross-sectional area was determined from at least 150 fibres taken from six different fields in each hemidiaphragm, using NIS software (Nikon Instruments Europe B.V., Amsterdam, The Netherlands). Graphpad Prism (Graphpad Software, San Diego, CA) was used to calculate and plot the means and standard errors of the mean of measured quantities. Differences between the mechanically ventilated and MV+stim conditions were assessed using paired t-tests and were considered significant at p,0.05. In the three animals studied, mechanically ventilated hemidiaphragms showed signs of severe damage consisting of hypercontracted fibres with apparent lipid droplet accumulation and intense oedematous infiltrate of the interstitium leading to fibre disorganisation (fig. 1a). These structural abnormalities were not observed in the MV+stim hemidiaphragms, even after 72 h of mechanical ventilation (fig. 1a). The LETTERS 280