Ultrasound performs better than radiographs.

We applaud the British Thoracic Society (BTS) for its efforts to improve patient care through scientific evidence. We thus recognise the recent guidelines on pleural procedures and thoracic ultrasound (TUS) as an important attempt to develop a rational approach to chest sonography. 1 However, we are concerned that the BTS has reached conclusions based on a less complete review of TUS. The guidelines state that 'the utility of thoracic ultrasound for diagnosing a pneumo-thorax is limited in hospital practice due to the ready availability of chest x-rays (CXR) and conflicting data from published reports'. 1 This conclusion appears to be based on a small (but landmark) study of 11 patients from 1986 to 1989, two small studies with only four pneumothoraces in one and another small series whose ultrasounds were retrospectively reviewed. Against these small and somewhat dated studies, a large number of recent investigations support a quite different conclusion. Many well-performed retrospective reviews and a number of prospective studies have compared TUS to chest radiographs (CXR) in the detection of pneumothoraces using CXR as the criterion standard. Noting the limitations of CXR in detecting pneu-mothoraces, we feel that only prospective studies utilising CT as the reference criterion are valid to assess the relative merits of ultrasound versus radiography. Although methodology and populations have varied, at least nine comparative trials, conducted in the last decade, have noted a higher sensitivity for TUS than CXR in the detection of pneumothorax. While the widely reported sensitivities (49%e100%) for TUS detection of pneumothoraces has not been explained, a more important point is that, in each of these studies, the sensitivity of TUS was significantly higher than CXR. Sonographic specificities were not significantly different from those of CXR, ranging from 94% to 100%. Furthermore, in the studies where it is reported, the likelihood ratios have ranged from 36 to 153. 2e4 Since a typical benchmark of a useful test is one that can generate positive likelihood ratios of greater than 10, these test characteristics have persuaded many, including the authors of two systematic reviews, that TUS is a more accurate test than supine anteroposterior CXR for the detection of pneumothorax. Finally, we would also like to take issue with the assumptions underlying the phrase 'ready availability of chest x-rays'. For many critical care and emergency department patients with sudden unexplained dyspnoea, the delay involved in obtaining a 'stat' portable CXR can be lethal. For …