Background
Bronchiolitis is a viral lower respiratory tract infection that affects children younger than 24 months and represents the leading cause of hospitalization in infants [
1]. The main responsible pathogen is Respiratory Syncytial Virus (RSV), with infection typically occurring as recurrent seasonal epidemics [
2,
3]. The treatment is primarily supportive and no specific etiological therapy is routinely used to limit the viral infection and reduce the severity of clinical course [
4,
5].
According to the most recent American Academy of Pediatrics guidelines [
4], the diagnosis of bronchiolitis is clinical and chest x-ray (CXR) should be reserved for severe cases in which signs of pulmonary complications are present or where the severity of respiratory effort leads to Intensive Care Unit (ICU) admission.
Nevertheless, there is high variation in use of diagnostic tests across hospitals and CXR is still performed in about 50% of bronchiolitis [
6‐
8], mainly to diagnose or rule out bacterial pneumonia. It has been shown that children with clinical bronchiolitis are more likely to receive antibiotics when radiography is performed owing to similar radiographic appearance of infiltrate and atelectasis [
9]. Moreover, even if in clinical practice alveolar infiltration is considered to be secondary to bacterial infection and bilateral interstitial infiltrates to atypical bacterial or viral infections, CXR is too insensitive to distinguish bacterial from viral pneumonia [
10]. Finally, radiographic images interpretation varies significantly among observers [
11].
Despite these well-known limitations, CXR is widely used in bronchiolitis, resulting in children exposure to ionizing radiations, increased medical costs, time spent, and potential complications due to unnecessary antibiotic prescription [
12,
13]. For this reason, in the last years many quality improvement methodologies have been attempted to minimize x-ray use in these patients [
14,
15]. Despite these attempts, no significant enhancement into clinical practice has been reached.
Lung Ultrasound (LUS) is a feasible, portable, easy to learn and non ionizing radiation technique. In the last decades it has become an emerging diagnostic tool for diagnosing pneumonia in adults and children, with remarkable sensitivity and specificity [
16‐
20]. Moreover, in the last years there has been great interest in using LUS to differentiate bacterial pneumonia from viral infections [
21,
22]. In this sense LUS may be the ultimate tool to diagnose or rule out bacterial pneumonia in children with clinical bronchiolitis and to identify who would benefit from antibiotics. Nevertheless, at present LUS is not included in the diagnostic work-up of bronchiolitis. In fact, although few studies describe the sonographic characteristics of bronchiolitis [
23,
24], none have investigated the role of LUS in children with clinical bronchiolitis and suspected pulmonary bacterial co-infection.
To our knowledge this is the first study to investigate the role of LUS in diagnosing pneumonia in children with acute bronchiolitis. The primary aim of this study was to assess the diagnostic accuracy and reliability of LUS for the detection of pneumonia in children with bronchiolitis and to evaluate the agreement between LUS and CXR in diagnosing pneumonia in these patients. Furthermore, we evaluated the interobserver agreement of LUS between a pediatric clinician and a pediatric radiologist who independently performed LUS.
Discussion
Bronchiolitis is the most common viral lower respiratory tract infection that affects children younger than 2 years [
1]. The diagnosis is clinical and guidelines currently do not recommend the routine use of CXR [
4]. However, CXR is still performed in a high percentage of cases, mainly to diagnose or rule out bacterial pneumonia requiring antibiotic treatment. The prevalence of pulmonary bacterial co-infection in bronchiolitis varies widely on literature, ranging from 9.7% [
31] to 42% in severe cases admitted to ICU [
32]. These variations may be due to several factors, including the age of children, the severity of bronchiolitis, the differences of assistance setting, the percentage of patients who underwent CXR, and the criteria to define CXR abnormal.
Our study investigated the reliability of LUS in discriminating children with uncomplicated bronchiolitis from those with concomitant bacterial pneumonia.
We observed a high rate of bronchiolitis complicated by bacterial pneumonia in our study (29%). This may depend on the highly selected setting of our patients, as we included only hospitalized children with bronchiolitis who underwent CXR because of the suspicion of concomitant bacterial pneumonia. The diagnosis of pneumonia was corroborated by the lower oxygen saturation (p 0.001) and higher CRP values (p 0.023) of these patients compared to those with uncomplicated bronchiolitis, together with their prompt response to antibiotic therapy.
In our study we found a strong correlation between CXR and LUS in diagnosing bacterial pneumonia (rs 0.64). Many recent studies reported similar findings, showing substantial agreement between the two techniques with kappa values of 0.64–0.89 [
33,
34]. Discordant results may be partially due to the superior sensitivity of LUS to detect subcentimetric consolidations [
16]. According to this, when only consolidation size > 1 cm was considered positive for pneumonia, the correlation in our study grew further (rs 0.68).
The good ability of LUS to diagnose pediatric pneumonia has been previously reported in literature [
16,
20,
35] but none of these studies focused on the setting of children with acute bronchiolitis. Consolidation has to extend to the pleural surface to be visualized by ultrasound and the supraclavicular region and/or the area covered by the scapula can be difficult to explore by LUS. However, studies on adults shows that pneumonia reach the pleura in 92% of hospitalized patients [
36] and up to 98% in the critically ill [
37]. A recent meta-analysis done on pediatric patients with suspected pneumonia [
18] revealed a higher sensitivity (96%, 95% CI 94–97%) but lower specificity (93%, 95% CI 90–96%) of LUS compared with adults data [
38]. The higher sensitivity may depend on the smaller thorax size and the thinner chest wall of children that lead to better visualization of the lung parenchyma by LUS [
36]. The lower specificity may be a result of non infiltrative processes including atelectasis, that are common in pediatric diseases as asthma or bronchiolitis and can be misinterpreted as pneumonia by ultrasound when the size of consolidation is small [
18].
According to these results, in our series we found a higher sensitivity of LUS compared to CXR (100% vs 96%) with a slightly lower specificity (84% vs 87%) when all sonographic consolidations with bronchogram were considered consistent with bacterial pneumonia. When only > 1 cm consolidations were considered positive, the specificity of LUS increased from 83.9 to 98.4%, raising a question about the diagnostic value of sonographic subcentimetric pneumonia.
These findings were consistent to those previously reported in literature [
16,
20,
35] that highlighted the uncertain pathological relevance of subcentimeter consolidations. LUS is generally useful for differentiating pneumonia from atelectasis caused by bronchial block to air entry [
26]. In pneumonia the bronchogram appears like a branching echogenic structure and can have intrinsic dynamic centrifugal movements due to breathing (“dynamic air bronchogram”) proving bronchial patency and ruling out atelectasis. On the contrary, the bronchogram in atelectasis has a parallel course and it is typically a “static air bronchogram” due to the absence of airflow secondary to the occluded airway [
26]. However the air bronchogram and its characteristics (parallel vs arborized, static vs dynamic) can be difficult to detect in subcentimeter lung consolidations. Subpleural consolidations - commonly present in bronchiolitis - can be misinterpreted as subcentimeter pneumonia by ultrasound. According to this, in our study LUS showed false-positive findings in 10 children, all but one consisting in subcentimeter pneumonia. LUS is without any radiation exposure and it requires short examination time (2–8 min) [
16] and lower cost compared to CXR [
20] as it can be performed by the pediatrician during the daily ward round with immediate bedside availability of results. As a result, we suggest that patients with subcentimeter consolidations should undergo ultrasound follow up before starting antibiotic therapy with the aim of identifying those who will achieve a spontaneous resolution. Moreover we evaluated the correlation between positive LUS (consolidations with bronchograms) and clinical/laboratory data (fever
> 38 °C, SatO2
< 92%, WBC
> 15,000/mmc, CRP
> 4 mg/dl). No strong correlation was founded. This may be due to the fact that no clinical manifestations nor laboratory markers are able to clearly differentiate bacterial from viral disease and predict severity of pediatric pneumonia [
39]. However a weak positive correlation emerged between positive LUS and SatO2
< 92%, CRP
> 4 mg/dl and TC
> 38 °C when all consolidations with bronchograms were included in the LUS positive findings. The positive correlation was confirmed between LUS findings and both SatO2
< 92% and CRP
> 4 mg/dl when considering only ultrasound consolidation size > 1 cm. If our results will be further confirmed in a larger populations in multicenter studies, a clinical and laboratory picture consistent with pneumonia, associated with positive LUS findings would exclude the need to perform a CXR.
To examine the effect of experience on LUS accuracy, we evaluated the interobserver agreement between a trained pediatrician and a pediatric radiologist sonologists in the first 30 patients enrolled in the study. We found an excellent interobserver agreement with a kappa value of 0.93, according to previous studies (kappa 0.55–0.93) [
16,
20,
40]. The distinctive feature of our study is that both users independently performed the ultrasounds, and this is important considering that LUS is an operator-dependent technique and incomplete chest exploration may cause diagnostic pitfalls. The emerged almost perfect interrater reliability (IRR) between novice and expert users supports that LUS is a basic easy-to-learn sonographic technique, as affirmed by the International Liaison Committee on Lung Ultrasound [
41].
Moreover we calculated the interobserver agreement of CXR between two pediatric radiologists on the same set of 30 patients to compare the reliability of CXR with LUS. We found a high IRR for CXR in detecting consolidation but lower than LUS (kappa 0.74 versus 0.93). Only one study did a similar comparison: 50 of the LUS and CXR images were read by 4 radiologists to calculate the interobserver agreement in diagnosing pediatric pneumonia, resulting in a poor IRR for CXR and moderate for LUS (kappa 0.36 versus 0.55) [
39]. Similarly, fair to moderate IRR for the interpretation of CXR for pediatric pneumonia is frequently reported in literature [
42,
43], even if it varies considerably depending on level of prior training of reporters [
44,
45]. In comparison with previous studies, the interobserver agreement of CXR in our study was unexpectedly high. This may have been because of the high experience level of our attending pediatric radiologists.
Our study had some limitations. First, this was a single center study with a relatively small numbers of children. Therefore, more studies with a larger sample size are required to confirm our data. The second limitation of the study is the lack of a true diagnostic reference standard due to ethical reasons. This is a common limitation in all studies assessing the diagnostic performance of LUS compared to CXR for the diagnosis of pneumonia in children. CT is normally considered the ideal gold standard for pneumonia, but it cannot be routinely used in children due to intensive exposure to radiations, availability and high cost. CXR is widely considered a crucial step in the diagnosis of pneumonia, nevertheless it is not 100% sensitive nor specific, and variation exists in intra- and interobserver agreement among radiologists [
46]. The limitations of CXR in the diagnosis of pneumonia are more evident in patients with bronchiolitis, because radiographic appearance of infiltrate is similar to atelectasis [
9] and CXR cannot reliably distinguish viral from bacterial pneumonia [
10,
47]. Thus, according to previous studies [
17,
48,
49], our diagnostic gold standard was the ex-post diagnosis of pneumonia made by an experienced pediatrician blinded to LUS findings, on the basis of clinical presentation, laboratory tests and CXR.
Despite these limitations, this study found useful results that support the reliability of LUS in evaluating pneumonia also in children affected by bronchiolitis, providing arguments for reducing CXR achievement.