The role of lung ultrasound as an alternative to routine chest roentgenograms for the detection of a pneumothorax after thoracic surgery is under discussion. Two large studies, which were performed under routine [16] or near-routine conditions [23], exhibited a low sensitivity (0.32, 0.19, respectively), whereas two smaller studies with rather artificial conditions (restrictive inclusion criteria [24], small sample size with very high pneumothorax rate [21]) revealed a 1.0 sensitivity. The reasons for this discrepancy are not clear; different sonographic examination techniques and inclusion criteria are suspected to be the cause.
Furthermore, the reported accuracy of postsurgical lung ultrasound is not consistent with the results of large meta-analyses from internal medicine and traumatology, where ultrasound was found to be superior to chest roentgenograms in the detection of a pneumothorax [13,14,15]. Potential reasons are 1) postoperative anatomic changes like mediastinal shift to the operated side, shrinkage of the operated hemithorax and pleural adhesions which do not regularly occur in medical or trauma patients and 2) that most X-rays in traumatology and intensive care medicine are shot in supine position, impairing the sensitivity of X-ray.
Our results demonstrate that “the usual suspects”, which are considered to debilitate thoracic ultrasound, do not influence the sensitivity of ultrasound under routine conditions. There are various hypotheses why an indwelling chest tube is associated with a higher sensitivity: First, the prevalence of a pneumothorax is higher on the first postoperative day than later. Second, a pneumothorax on the first day after surgery will be larger than later, otherwise one would not have decided to remove the chest tube. We have demonstrated that the sensitivity of ultrasound is dependent on the pneumothorax size [16]. Third, the examiner who – consciously or not – perceives an air leakage through a chest tube will probably expect a pneumothorax and examine more accurately (examiner bias / diagnostic suspicion bias [30, 31]). One should not assume a causal relationship, but also not consider a chest tube as an obstacle to lung ultrasound.
One reason for the wrong assumptions on the prerequisites of lung ultrasound seems to be an insufficient methodology. Most studies, also in internal medicine, assess only or mainly lung sliding to detect or rule out a pneumothorax. While lung sliding is completely specific against a pneumothorax, the inverse conclusion is not true. Absence of lung sliding is found in pleurodesis, fibrosis, COPD, shallow breathing patients (for example after thoracic surgery due to pain), pulmonary emphysema and other pathologic conditions [32]. Relying on lung sliding only leads to a higher false positive rate.
Lung pulse is a powerful, but often neglected artefact to rule out pneumothorax in difficult situations [16, 33].
The addition of B-lines and consolidations increases the accuracy in ruling out a pneumothorax, however, they are not exhaustive in some cases, and there are severe discrepancies in the terminology of “B-lines” and “comet-tail” artefacts [34, 35]. One should be aware of comet tails which mimic B-lines; for example, E-lines are generated by a subcutaneous emphysema and could let an uncareful examiner unjustly rule out a pneumothorax. An insufficient discrimination of different types of comet tails leads to a higher false negative rate. This is consistent with our data, with subcutaneous emphysema being the only covariate which significantly impaired the specificity of lung ultrasound, although we were absolutely aware of the pitfalls of subcutaneous emphysema. The physician who uses ultrasound should indicate radiologic imaging generously if the examination is contradictory or impaired for example by severe skin emphysema.
The reliable detection of a pneumothorax by the lung point can be difficult after thoracic surgery due to altered anatomy or impaired access to the entire hemithorax. In a large pneumothorax, the lung point is not present. Thus, ruling out a pneumothorax by sonography is easier than its detection.
If ultrasound rules out a pneumothorax, no further radiologic imaging is necessary to answer this question, due to the high negative predictive value of ultrasound for a clinically relevant pneumothorax. In contrast, if ultrasound is inconclusive or contradictory to clinical findings, radiologic imaging is obligate. As chest X-ray may also be altered by supine position, obesity, hypoventilation, severe emphysema, etc. in some of these cases a more precise – and more cumbersome – technique is required, which is computed tomography. From the authors’ point of view, ultrasound is a fast and cost-effective tool at the low end of the diagnostic ladder, but also needs training to unfold its full potential.
Limitations
Taking standard chest roentgenograms as the reference test impairs the calculation of diagnostic accuracy. Staquet et al. [20] described that the sensitivity of a diagnostic index test (ultrasound) can be calculated exactly, if the reference test (roentgenogram) is a hundred per cent specific, which is the case, indeed [13,14,15]. However, we can only calculate the lower bound of the specificity range. Former trials which were controlled against computed tomography showed that ultrasound is nearly 100% specific for a pneumothorax [19]. Thus, most of “false positives” in the index test are real pneumothoraxes which were overseen by X-ray, i.e. false positives of the reference test. Therefore, we assume that the sensitivity of chest ultrasound should be higher if tested against a perfect reference test (computed tomography).
However, this drawback is the same for every subgroup. Therefore, our data help to find the correct parameters and appropriate patients for ultrasound examination [36]. If forthcoming trials are performed under wrong assumptions, lung ultrasound in thoracic surgery will remain an academic tool for a niche. Furthermore, for certain clinical applications, sensitivity is not the most important test parameter. Two studies have reported the usefulness of lung ultrasound despite its low sensitivity [22, 37]. For instance, the negative predictive value for relevant pneumothorax sizes is clinically more important after chest tube removal, since small apical pneumothoraces are usually not of clinical interest [16, 22].
Particularly after chest tube removal, routine roentgenograms rarely have any consequences (0.9% in 1097 patients, [1]). Bjerregaard et al. therefore recommend an individualised approach based on clinical information. Ultrasound could be such a tool to significantly reduce the number of postoperative roentgenograms.