Presence of a nasogastric tube is a risk factor for the development of ventilator-associated pneumonia (VAP). Alternatively, gastrostomy can be used for administration of enteral feedings. To determine whether early performance of gastrostomy affects frequency of VAP, a randomised, controlled study was carried out in patients mechanically ventilated for stroke or head injury. In the gastrostomy group, patients underwent the procedure within 24 h of intubation. A nasogastric tube was inserted in controls. Individual subjects were studied for 3 weeks. In total, 20 subjects (mean age 48+/-15.2 yrs) were allocated to the gastrostomy group, and 21 to the control group (46.6+/-15.4 yrs). Of these groups, two (10%) and eight (38.1%) developed VAP, respectively. Four patients with gastrostomy and three controls did not complete the study (due to weaning from ventilatory support or death). After excluding these subjects, difference in VAP frequency persisted: two out of 16 subjects with gastrostomy had VAP (12.5%) versus eight out of 18 controls (44.4%). There were no differences in duration of hospitalisation or mortality between the two groups. In conclusion, in patients mechanically ventilated for stroke or head injury early gastrostomy is associated with a lower frequency of ventilator-associated pneumonia compared with a nasogastric tube.