The diagnostic accuracy of chest ultrasound for CT-detected radiographic consolidation in hospitalised adults with acute respiratory failure: a systematic review

Abstract

Objectives: (1) Summarise chest ultrasound accuracy to diagnose radiological consolidation, referenced to chest CT in patients with acute respiratory failure (ARF). (2) Directly compared ultrasound with chest X-ray.

Setting: Hospitalised patients.

Participants: Studies were eligible if adult participants in respiratory failure underwent chest ultrasound to diagnose consolidation referenced to CT. Exclusion: (1) not primary study, (2) not respiratory failure, (3) not chest ultrasound, (4) not consolidation, (5) translation unobtainable, (6) unable to extract data, (7) unable to obtain paper. 4 studies comprising 224 participants met inclusion.

Outcome measures: As planned, paired forest plots display 95% CIs of sensitivity and specificity for ultrasound and chest X-ray. Sensitivity and specificity from each study are plotted in receiver operator characteristics space. Meta-analysis was planned if studies were sufficiently homogeneous and numerous (≥4). Although this numerical requirement was met, meta-analysis was prevented by heterogeneous units of analysis between studies.

Results: All studies were in intensive care, with either a high risk of selection bias or high applicability concerns. Studies had unclear or high risk of bias related to use of ultrasound. Only 1 study clearly performed ultrasound within 24 h of respiratory failure diagnosis. Ultrasound sensitivity ranged from 0.91 (95% CI 0.81 to 0.97) to 1.00 (95% CI 0.95 to 1.00). Specificity ranged from 0.78 (95% CI 0.52 to 0.94) to 1.00 (0.99 to 1.00). In two studies, chest X-ray had lower sensitivity than ultrasound, but there were insufficient patients to compare specificity.

Conclusions: Four small studies suggest ultrasound is highly sensitive and specific for consolidation in ARF, but high risk of bias and concerns about applicability in all studies may have inflated diagnostic accuracy. Further robustly designed studies are needed to define the role of ultrasound in this setting.

Trial registration number: http://www.crd.york.ac.uk/PROSPERO/ (CRD42013006472).

Figure 1

PRISMA flow diagram of study identification and selection.

Figure 2

QUADAS2 (Quality Assessment of Diagnostic Accuracy Studies) risk of bias and applicability assessment of individual studies.

Figure 3

QUADAS2 (Quality Assessment of Diagnostic Accuracy Studies) risk of bias and applicability assessment across primary studies.

Figure 4

Sensitivity and specificity of ultrasound and chest X-ray for included studies. Note different units of analyses which preclude pooling of studies: lung regions (12 per patient) in Lichtenstein et al 2004; lungs (2 per patient) in Lichtenstein et al 2004 and Xirouchaki et al 2011; and individual patients in Refaat, Abdurrahman 2013.

Figure 5

Sensitivity and specificity of ultrasound and chest X-ray for consolidation.

Figure 6

Direct comparisons for ultrasound and chest X-ray in two individual studies. Units of analysis are lungs (2 per patient) for Xirouchaki et al 2011, and lung regions for Lichtenstein et al 2004 (12 per patient).

Figure 7

Impact of units of analyses on sensitivity and specificity of ultrasound and chest X-ray for consolidation. Data from Xirouchaki et al 2011 are stratified according to two units of analysis; ‘lung’ (2/patient) and ‘lung region’ (12/patient). In comparison to lung analysis, lung region analysis reduces sensitivity but inflates specificity. It also increases total study numbers, giving the appearance of tighter estimates of precision.