Systematic review and meta-analysis of the accuracy of lung ultrasound and chest radiography in diagnosing community acquired pneumonia in children
Chenxi Shi,Xinmin Xu,Yongsheng Xu
DOI: https://doi.org/10.1002/ppul.27221
2024-09-06
Abstract:Chest radiography (CXR) is commonly used for diagnosing childhood pneumonia, but concerns about radiation exposure have raised interest in using radiation-free lung ultrasound (LUS) as an alternative imaging modality. Therefore, we designed this meta-analysis to compare the accuracy of LUS and CXR for diagnosing childhood pneumonia. We searched 8 databases and 1 clinical trial registry for studies published from inception to March 2023. Studies assessing lung ultrasound and chest radiography for diagnosing childhood pneumonia were included. Two reviewers independently screened literature, extracted data, and assessed the risk of bias using the QUADAS-2 tool for each study. Meta-analysis was conducted using a random-effects model, and pooled sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, diagnostic odds ratio, and summary receiver operating characteristic (SROC) curve were assessed. Statistical analyses were performed using Meta-Disc 1.4, RevMan 5.4, and Stata 17.0 software. Heterogeneity was examined, and subgroup analysis was conducted to explore the accuracy of lung ultrasound in diagnosing childhood pneumonia. Out of the 4089 screened articles, 30 studies were included, encompassing a total of 4546 children. Of those, 3257 were diagnosed with pneumonia, 3190 through LUS, and 2925 via CXR. The meta-analysis showed that the sensitivity, specificity, positive and negative likelihood ratios, and diagnostic odds ratio of LUS were 0.940 (95% CI 0.930-0.949), 0.855 (95% CI 0.835-0.873), 7.561 (95% CI 4.956-11.536), 0.08 (95% CI 0.056-0.113), and 110.77 (95% CI 62.156-197.40), respectively. The combined area under the SROC curve was 0.9712, Q index = 0.9218. For CXR, the sensitivity, specificity, positive and negative likelihood ratios, and diagnostic odds ratio were 0.893 (95% CI 0.881-0.905), 0.906 (95% CI 0.889-0.921), 18.742 (95% CI 7.551-46.520), 0.105 (95% CI 0.062-0.180), and 237.43 (95% CI 74.080-760.99), respectively. The combined area under the SROC curve was 0.9810, Q index = 0.9391. Subgroup analysis showed that the implementation location, interval between lung ultrasound and chest radiography, and operator experience had no impact on the accuracy of lung ultrasound in diagnosing childhood pneumonia. Existing evidence suggests that lung ultrasound has high accuracy for diagnosing childhood community-acquired pneumonia. Compared with chest radiography, lung ultrasound has higher sensitivity, similar specificity, and advantages such as radiation-free, lower cost, simplicity of operation, and ease of follow-up, making it an important imaging modality for diagnosing childhood pneumonia.