The 'cuff-leak' test, which involves demonstrating a leak around a tracheal tube with the cuff deflated, has been advocated to determine the safety of extubation in patients with upper airway obstruction. In 62 such patients we were able safely to extubate all patients with a cuff leak. Two patients extubated without cuff leak required reintubation and in five patients who repeatedly failed the test, tracheostomy was performed. Subsequently, we extubated 10 patients who were stable on spontaneous ventilation and did not have cuff leak; three later required tracheostomy and seven were uneventfully extubated. While the presence of cuff leak demonstrates that extubation is likely to be successful, a failed cuff-leak test does not preclude uneventful extubation and if used as a criterion for extubation may lead to unnecessarily prolonged intubation or to unnecessary tracheostomy.