Sensitivity of lung ultrasound for the detection of pneumothorax one day after pulmonary resection—a prospective observational study

Thoracic surgery is usually followed by numerous chest X‑rays (CXR). They are associated with high logistic efforts and patient discomfort, are binding multiple personnel resources (transport, technical assistants, radiologists), are costly, and have very limited diagnostic and therapeutic value [1]. Nonetheless, they represent the clinical standard, and there are no recommendations on the appropriate use of chest X‑rays in the postoperative course [2]. In the last two decades, lung ultrasound (LU) has been increasingly used to identify pneumothoraces and other pulmonary pathologies, exhibiting test parameters, which are superior to CXR and compete with computed tomography (CT) (sensitivity/specificity LU 68–91%/97–100%, CXR 28–76%/100%, [3‐5]). However, these findings arise from traumatology, spontaneous pneumothorax, and ventilated intensive care patients who differ from thoracic surgery patients. Postoperative conditions, such as reduced lung volume, diaphragm displacement, mediastinal shift, and adhesions can alter the distribution of extra-pulmonary air within the chest and impair the sonographic accessibility. Hence, data on the accuracy of lung ultrasound in postoperative patients is limited. Reported cohorts are heterogeneous regarding type of surgery, patient selection, prevalence of a postoperative pneumothorax, positioning of the ultrasound probe, and time of the ultrasound examination (Table 1). Thus, reported sensitivities of lung ultrasound vary widely from 21 to 100% (Fig. 1; [6‐8]). There is a need for further trials on postsurgical patients under well-defined conditions. We conducted a prospective trial on patients with pulmonary surgery, as these patients have the highest risk for a postoperative pneumothorax, using rigorous real-life conditions.

A total of 68 consecutive patients were included; their basic characteristics are given in Table 2.

The median time between ultrasound and X‑ray was 80 min (interquartile range 42.5, 101.75). Most X‑rays (91%) were shot in supine position. The mean duration of sonographic examination was 145 ± 64 s.

Using ultrasound, a pneumothorax was identified in 18 patients and ruled-out in 46 patients. In 4 patients, diagnosis was not possible due to extensive soft tissue emphysema: 2 patients were CXR-negative, 2 patients were CXR-positive for a small pneumothorax (1.4 and 0.6 cm, respectively). The signs by which a pneumothorax was detected or ruled-out are presented in Table 3.

Using X‑ray, 23 pneumothoraces were found; the mean size was 1.5 ± 1.0 cm.

The distribution of sonographic and radiologic findings of pneumothorax after removing the chest tube is presented in Table 4. The observed prevalence of a pneumothorax was 0.34.

The diagnostic test parameters are

A therapeutic conformity was observed in 96% of cases. In three cases, the sonographic recommendation was to remove the chest tube, whereas it was actually retained. Of these 3 patients, two had a small, clinically irrelevant pneumothorax of 1.0 cm and 1.8 cm in CXR, respectively.

On the first postoperative day, 30% of patients had air leakage via the chest tube. Air leakage was present in 78% of sonographically identified pneumothoraces, but only in 48% of radiologically identified pneumothoraces. There was no air leakage in patients who were negative for pneumothorax in both ultrasound and X‑ray. Air leakage correlated weakly with X‑ray (spearman’s rho = 0.26) and moderately with ultrasound (spearman’s rho = 0.43).

To our knowledge, this is the first study on the postoperative use of LU for pneumothorax, which was performed on unselected patients under real-life conditions and used a rigorous sonographic methodology.

Nonetheless, we found a sensitivity of LU for postoperative pneumothorax which is in between the previously reported data in noncardiac thoracic surgery, whereas the sensitivity of CXR was in the known range. The sensitivity of LU was only slightly lower than that of CXR (0.48 vs. 0.56). Similar to previous studies [6, 7], our population contained only operations with a considerable risk for postoperative pneumothorax. However, Goudie et al. excluded unconscious and physically weak patients because patients were examined in sitting position, and counted each hemithorax individually in the statistical analysis, which mathematically explains their low prevalence of pneumothorax (Table 1). They also excluded patients with severe subcutaneous emphysema or wound dressings. Although we did not have these restrictions in our study, we found a better sensitivity.

The trial of Patella et al., which reported excellent test parameters for LU, had very selective inclusion criteria [15] and another time of examination (after chest tube removal). The study of Smargiassi et al., who also demonstrated a high sensitivity of LU and an unusually high prevalence of pneumothorax (83%), did not declare the type of surgery [8]. Thus, the reasons for the variation in the performance of LU remain unclear.

Apart from patient selection criteria, the sonographic technique and terminology may be major factors contributing to varying test results. Several LU studies (e.g. [6, 7, 16]) consider the absence of lung sliding and comet tails as a pneumothorax, which may overestimate the prevalence of pneumothorax. However, the absence of lung sliding is also found in reduced lung compliance, fibrosis, bullous emphysema, and is therefore not specific [15, 17]. “Comet tails” is a collective term that includes some artifacts which rule out a pneumothorax (B-lines, I‑lines) and some which do not (e.g. E‑lines, Z‑lines) [11]. So if a report uses the term “comet tail”, the reader cannot be sure if a reliable method was used [18]. The validity increases when additional patterns are sought, like lung-pulse and consolidations.

Our study, like all others, is limited by the lack of a perfect reference test, i.e. CT. Ethically, it is unacceptable to perform a CT to just assess the presence of postoperative pneumothorax. The sensitivity of X‑ray is limited, as most patients are examined in supine position on the first day after surgery in daily routine. However, to determine the role of ultrasound as a potential diagnostic alternative, LU must compete with the accuracy of CXR under routine conditions rather than with CT.

Given the lack of CT findings, the pneumothorax size could not be reliably determined. A valid (semi)quantification is not possible with CXR in supine position. Hence, it is unclear whether sonographically false-negative pneumothoraces were of relevant size or not. Nonetheless, the therapeutic conformity between sonographic and conventional assessment was very high, i.e. that the patient would have been treated equally by either LU or CXR. It seems that a negative LU is sufficiently reliable in excluding a pneumothorax, which is represented by the decent negative predictive value in this study. For the clinical question of a relevant pneumothorax, the NPV of LU is more important than its sensitivity. Considering that less than 1% of postoperative CXR have a therapeutic consequence [1], it is the task of future studies to investigate the clinical relevance of LU in a large-scaled population, rather than just sensitivity.

To represent the actual clinical routine, we decided to avoid restrictive patient selection. Inherently, this could have resulted in lower sensitivity, due to the inclusion of patients with impaired sonographic exam conditions, such as obesity or COPD.

Our results are relevant particularly for, but not limited to pulmonary surgery. Cardiac and oesophageal surgery use chest tubes as well, and there are efforts in these areas to reduce radiation according to the ALARA principle (as low as reasonably achievable), too [19, 20]. However, since the postoperative pneumothorax is a distinct entity pathophysiologically different from a spontaneous or traumatic pneumothorax, the experiences from (a) nonsurgical trials or (b) surgical trials with artificial, “academic” conditions should not be applied carelessly to routine surgical patients. Instead, we need large-scale studies which respect the particular anatomy and physiology of the postoperative chest. Our data can serve for planning the sample size of these future studies.

In conclusion, we found a moderate sensitivity of LU for postoperative pneumothorax in an unselected routine patient population, nearly reaching the sensitivity of CXR. Although the sensitivity of LU was slightly lower than CXR, therapeutic decisions based on LU would have resulted in equally safe patient treatment. For further controlled studies, it may be reasonable to introduce a stepwise approach with ultrasound followed by CXR on demand in the case of inconsistent findings or poor exam conditions. This approach could reduce the use of CXR after thoracic (pulmonary, cardiac, or oesophageal) surgery.