Role of Lung Ultrasound in the Detection of Lung Sequelae in Post-COVID-19 Patients: A Systematic Review and Meta-Analysis

Background: During the COVID-19 pandemic, several studies demonstrated the effectiveness of lung ultrasound (LUS) as a frontline tool in diagnosing and managing acute SARS-CoV-2 pneumonia. However, its role in detecting post-COVID-19 lung sequelae remains to be fully determined. This study aims to evaluate the diagnostic accuracy of LUS in identifying lung parenchymal damage, particularly fibrotic-like changes, following COVID-19 pneumonia, comparing its performance to that of CT. Methods: Relevant studies published before July 2024 were identified through a comprehensive search of PubMed, Embase, and Cochrane library. The search terms were combinations of the relevant medical subject heading (MeSH) terms, key words and word variants for “lung”, “post-COVID”, “long-COVID”, and “ultrasound”. The pooled sensitivity, specificity, diagnostic odds ratio (DOR), and summary receiver-operating characteristic (SROC) curve were used to examine the accuracy of CEUS. The selected works used different thresholds for the detection and counting of B-lines by ultrasound. This led to dividing our analysis into two models, the first based on the lower thresholds for detection of B-lines found in the works, and the second on data obtained using a higher detection threshold. Results: In terms of the diagnostic accuracy of LUS in detecting residual fibrotic-like changes in patients post-COVID-19 infection, a low-threshold model displayed a pooled sensitivity of 0.98 [95% confidence interval (CI): 0.95–0.99] and a pooled specificity of 0.54 (95% CI: 0.49–0.59). The DOR was 44.9 (95% CI: 10.8–187.1). The area under the curve (AUC) of SROC was 0.90. In the second analysis, the model with the higher threshold to detect B-lines showed a pooled sensitivity of 0.90 (95% CI: 0.85–0.94) and a pooled specificity of 0.88 (95% CI: 0.84–0.91). The DOR was 50.4 (95% CI: 15.9–159.3). The AUC of SROC was 0.93. Conclusions: In both analyses (even using the high threshold for the detection of B-lines), excellent sensitivity (98% in model 1 and 90% in model 2) is maintained. The specificity has a significant variation between the two models from 54 (model 1) to 87% (model 2). The model with the highest threshold for the detection of B-lines displayed the best diagnostic accuracy, as confirmed by the AUC values of the SROC (0.93).