Lung Ultrasonography Provides Robust, Predictive Clinical Data About COVID-19

Lung ultrasound (US) parameters have high discriminative accuracy across a range of clinical settings for COVID-19, according to study results published in The Journal of Infectious Diseases.

Investigators from the RAISON and EPICC research groups conducted an international multisite study to assess the generalizability and utility of lung US for COVID-19 prognostication across various clinical settings. Between June 2020 and June 2023, adults with polymerase chain reaction (PCR)- or rapid antigen-confirmed COVID-19 were prospectively enrolled in 5 cohorts designated for validation at Johns Hopkins Hospital (n=75; 39.3%), Duke University Hospital (n=28; 14.7%), Walter Reed National Military Medical Center (n=42; 22.0%), and Madigan Army Medical Center (n=15; 7.9%) in the United States, as well as Fort Portal Regional Referral Hospital (n=33; 16.2%) in Uganda. Data from an additional 264 patients from Johns Hopkins Hospital were used as the derivation cohort. At enrollment, patients underwent 12-zone lung US and 6-second clips were read offsite. The primary outcome was progression to higher-level care on the basis of lung US results. Logistic regression models were used in the analyses.

The pooled validation vs derivation cohort was more predominately women (57.9%; P =.002) and younger (median age, 45.0; P <.001), and comorbid conditions such as hypertension (35.1%), chronic pulmonary disease (16.8%), and diabetes (15.7%) were less common. At baseline, 84 (44.9%) patients in the validation cohort were hospitalized. Progression in oxygen requirements, hospitalization, or death during the study were observed in 11 (5.7%) patients.

In the validation cohort, the top 5 predictors for future progression were percent of lung fields with B-lines (cross-validated area under the receiver operating characteristic curve [cvAUC], 0.88; 95% CI, 0.87-0.90), discrete B-lines (cvAUC, 0.87; 95% CI, 0.85-0.88), oxygen saturation (cvAUC, 0.82; 95% CI, 0.81-0.84), A-lines (cvAUC, 0.80; 95% CI, 0.78-0.81), and mean lung US score (cvAUC, 0.74; 95% CI, 0.72-0.76).

The top 5 predictors in the derivation cohort were respiratory rate (cvAUC, 0.7; 95% CI, 0.68-0.72), percent of lung fields with A-lines (cvAUC, 0.66; 95% CI, 0.65-0.67), mean lung US score (cvAUC, 0.63; 95% CI, 0.61-0.66), B-lines (cvAUC, 0.63; 95% CI, 0.60-0.66), and confluent B-lines (cvAUC, 0.63; 95% CI, 0.61-0.65).

Using a specificity cutoff of 70%, observing A-lines in less than 1/4 of lung fields (78%; 95% CI, 51%-100%) or B-lines in 1/2 and more of lung fields (78%; 95% CI, 51%-100%) was sensitive for predicting progression in the validation cohort. B-line sensitivity was similar between the 2 cohort groups, while specificity with any B- or discrete B-lines was higher in the validation vs derivation cohort.

In a pooled population of both the validation and derivation cohorts (n=454), the best predictors for COVID-19 progression were percent of B-lines (cvAUC 0.73; 95% CI, 0.72-0.73), percent of A-lines (cvAUC 0.71; 95% CI, 0.71-0.72), respiratory rate (cvAUC 0.70; 95% CI, 0.69-0.71), percent of discrete B-lines (cvAUC 0.70; 95% CI, 0.69-0.70), and mean lung US score (cvAUC 0.69; 95% CI, 0.68-0.69).

The best performing composite model comprised the predictors percent of B-lines, respiratory rate, heart rate, and percent of A-lines (cvAUC, 0.78; 95% CI, 0.77-0.80).

Study limitations include the heterogenous study populations, small sample sizes, and the inability to obtain arterial gas measurements.

“Validation of ultrasound models within target populations is critical for determining the clinical utility prior to adoption,” the investigators concluded.

Disclosure: One study author declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of authors’ disclosures.

This article originally appeared on Infectious Disease Advisor