BACKGROUND: Complicated (recurring or persistent) spontaneous pneumothorax requires treatment either by talc pleurodesis with bullae electrocoagulation or, more aggressively, by thoracotomy or video-assisted thoracoscopic surgery. However, the relative merits of bullectomy, pleurectomy, and pleurodesis have not yet been established in the treatment of spontaneous pneumothorax. METHODS: The complications, duration of drainage, length of hospital stay, and immediate and long term success rate of treating complicated spontaneous pneumothorax with talc pleurodesis under local anaesthesia supplemented with nitrous oxide were studied. RESULTS: Talc pleurodesis was performed in 93 patients without serious complication (two benign arrhythmias, two subcutaneous emphysema, two pneumonia, one bronchospasm). The procedure was immediately successful in 90 patients (97%) with a median duration of drainage of five days (range 2-40) and a median length of hospital stay of 5.2 days (range 3-40). After a mean follow up duration of 5.1 (range 1-9.4) years in 84 cases the long term success rate was 95%, although six cases developed a small localised recurrence of spontaneous pneumothorax which did not require further surgery. Macroscopic staging at thoracoscopy was only carried out in the last 59 cases of whom 10 (17%) had bullae with a diameter of > 2 cm. In this group of patients the risk of definitive failure requiring surgery was significantly higher than in those patients without such bullae (odds ratio 7; confidence interval 3.7 to 13.3; p = 0.03), although eight of these patients did not require thoracotomy. Total lung capacity was reduced immediately after talc pleurodesis (mean (SD) 75 (23)% predicted at 10 days) but had improved to 95 (14)% predicted at 12 months. CONCLUSIONS: This study shows that simple thoracoscopic talc pleurodesis under local anaesthesia is a safe and effective treatment for complicated spontaneous pneumothorax. However, patients with bullae of > 2 cm in diameter have a greater risk of treatment failure.