Background
By favoring a more homogeneous distribution of tidal volume and inducing the recruitment of dorsal areas of lungs, prone positioning improves oxygenation, lung compliance, and ventilation/perfusion matching [
1,
2]. In selected patients with acute respiratory distress syndrome (ARDS) on invasive mechanical ventilation, prone positioning has been shown to reduce mortality [
3]. As a noninvasive and low-cost treatment, awake prone positioning (APP) has been extensively used in non-intubated patients with coronavirus disease (COVID-19) induced acute hypoxemic respiratory failure (AHRF) [
4]. We recently found that APP reduced the intubation rate within 28 days of enrollment for COVID-19 patients treated by high-flow nasal cannula (HFNC) [
5]. However, whether or not the better lung aeration with APP can explain the better patient outcome in terms of intubation needs has been poorly investigated.
Lung ultrasound (LUS), a noninvasive, radiation-free, and bedside-available imaging tool, has gained popularity for lung assessment in critically ill patients [
6]. Several studies have shown that LUS is a valuable tool in assessing lung aeration response to prone positioning in intubated ARDS patients [
7‐
11]. However, the ability of LUS to predict patient response to prone positioning was variable between studies [
7‐
10]. More recently, LUS has been used to assess the response to APP in non-intubated patients with COVID-19. APP responders [
12], who were defined by the increment of partial pressure of arterial oxygen/fraction of inspired oxygen [PaO
2/F
IO
2] ≥ 20 mmHg, had a greater decrease in LUS score after 3 h of APP than non-responders. However, the included patients had mild-to-moderate AHRF, the patients’ response was investigated only after the first session, and the predictors of intubation were not assessed due to the small sample size and the low intubation rate.
In our previous cohort study of prone-positioning in intubated patients with COVID-19, we found that although most patients had an improvement in oxygenation on the first session of prone positioning, only those who survived had a significant response over the second and third sessions of prone positioning, which suggests that serial measurements along the first three days of treatment rather than the first session only, might improve the accuracy of prediction regarding patient-centered outcomes [
13]. Moreover, in a large randomized controlled trial investigating the effects of APP on non-intubated patients with COVID-19, we found that LUS scores after three days of APP treatment decreased only in patients who eventually avoided intubation/death [
14]. Therefore, we performed this study with two aims: 1) to explore the aeration response to APP in patients with AHRF induced by COVID-19 by using LUS within the first three days, and 2) to explore whether the changes in LUS associated with APP can predict the need for intubation. We hypothesized that better lung aeration with APP was associated with a reduction in intubation rate.
Methods
Study design
This prospective observational study was registered on clinicaltrials.gov (NCT04855162), and performed at two hospitals with IRB approval (Rush University Medical Center No. 21040601 and Comité de Ética en Investigación Hospital Civil Fray Antonio Alcalde No. HCG/CEI-0753/21). Due to the noninvasive nature of the assessment, written consent was waived by ethics committees at both hospitals.
Patients
Consecutive patients ≥ 18 years with AHRF induced by COVID-19 confirmed by RT-PCR, treated by HFNC and APP were included. AHRF was defined by pulse oximetry (SpO
2)/FiO
2 < 315. Pregnant patients and those on palliative care or extracorporeal membrane oxygenation were excluded. Patients with APP ≤ 1 h/day or any missed LUS on the first APP session within the first three days of enrollment were withdrawn. Patients were encouraged to stay in the prone position for ≥ 8 h/day, as in our meta-trial, the treatment success was related to this threshold [
5].
Study procedures
LUS was performed by clinicians with > 8 years of clinical experience in LUS on critically-ill patients and certified trainers by WINFOCUS (
World Interactive Network Focused on Critical Ultrasound). An Edge II (Fujifilm) or TE7 ultrasound system (Mindray, Shenzhen, China) with a curved transducer (3–8 MHz) was used. Depth was set at 6–10 cm according to patient size, with focus placed in pleural line, and gain regulated to optimize lung artifacts. The LUS evaluations were performed daily within the first three days of enrollment and on the morning of the 4th day before APP, as long as patients had not been intubated. The LUS investigation was done 5–10 min before APP (pre-APP), 1 h after APP (post-APP), and 5–10 min after returning to the supine position (post-supine) following the first prone session of these days. At the supine position, dorsal zones were scanned on a short transient lateral decubitus position. Global LUS aeration score was measured over 12 lung zones (ventral, lateral, and dorsal), with total scores ranging from 0 to 36, with higher scores indicating less lung aeration. We chose this aeration score rather than the re-aeration score (from − 5 to + 1) [
15], because our study involved comparisons between groups in the pre-prone state more than once, with changes monitored over three days rather than a single APP session. Moreover, it is strongly correlated with tissue density assessed by CT scan and extravascular lung water assessed by transpulmonary thermodilution [
15,
16]. Scores were interpreted and validated by two expert clinicians blinded to the LUS procedure. SpO
2/FiO
2 ratio, respiratory rate (RR), and ROX index ([SpO
2/FiO
2]/RR) [
17,
18] were also recorded at the same time points of LUS assessment. Criteria for intubation were standardized and similar to our published meta-trial [
5].
Outcomes
The primary outcome was the change in global LUS score from pre-APP on day 1 to pre-APP on day 4. Secondary outcomes included changes in SpO2/FiO2 ratio, RR, and ROX index, as well as the differences in these variables between patients who were intubated (treatment failure) and those who avoided intubation (treatment success), and between responders and non-responders. Responders were defined as those patients whose SpO2/FiO2 ratio increased by ≥ 20% after the supine positioning of the first APP session. Predictors of treatment success were also explored.
Sample size
Using a mean of pre-prone LUS score of 18.7 and standard deviation (SD) of 4.4 and post-supine LUS score at day 3 of 16.9 and SD of 4.6 [
14], a confidence level (1-α) of 95% and power (1-ß) of 95%, the number of patients was 70. Considering an attrition rate of 5%, we calculated the total sample size as 74.
Statistical analysis
The normality of distribution for continuous variables was assessed by Kolmogorov–Smirnov test, and presented as mean ± SD or as median and interquartile range (IQR). Repeated measures ANOVA or Friedman’s test was used to compare differences of the variables pre-APP, post-APP, and post-supine. Comparison of continuous variables between groups (treatment success vs treatment failure, responders vs non-responders) were conducted with Student’s t or Mann–Whitney U test, while ANCOVA test was used to compare variables between two groups at the same time points with Bonferroni adjustment for baseline covariates. Correlation coefficients were assessed with Spearman’s test. Categorical variables are presented as counts and proportions with 95% confidence intervals (95% CIs) and were compared by Chi-square or Fisher’s exact test. To explore if the LUS score was associated with treatment success, multivariate logistic regression models were performed with one covariate introduced at a time to avoid overfitting, including those with a p-value < 0.20 in the univariate comparisons between groups. The accuracy of different variables in predicting treatment success was assessed by calculating the area under the receiver operating characteristic (AUROC) curves. Two-sided p ≤ 0.05 was considered statistically significant. Statistical analysis was performed with MedCalc (MedCalc Software Ltd Ostend, Belgium. Version 20.1) and GraphPad Prism software (version 9.3.1).
Discussion
In this study among patients with COVID-19-induced AHRF, we found that: 1) compared to patients with treatment failure, those with treatment success had a greater reduction in dorsal LUS score post-supine following the first APP session, 2) changes of RR, ROX index, SpO2/FIO2 after the first APP session did not differ between patients with treatment success vs. failure, 3) a reduction in the dorsal LUS score > 1 after the first APP session was independently associated with a lower risk of treatment failure, and 4) mean daily APP duration was significantly correlated to a decrease in global LUS score, and particularly to the reduction in dorsal LUS score for patients with treatment success.
Franchineau and colleagues reported time-dependent effects of prone positioning on lung recruitment assessed by electrical impedance tomography (EIT) in intubated ARDS patients on veno-venous ECMO, especially in dorsal zones [
19]. This time-dependent improvement in the aeration of the dorsal zones may lead to a more homogeneous distribution of lung inflation, improving lung compliance, decreasing the inspiratory effort [
20], and providing a more protective distribution of stress and strain.
Furthermore, we also found that the reduction in dorsal LUS score > 1 in the first APP session predicted the treatment success. Importantly, this finding can help early identify those patients with high risk of APP failure, who might benefit from more intensive monitoring and early treatment escalation and not delay intubation [
21]. Thus, LUS assessment may be a very useful tool in the day-to-day clinical decision-making process.
In contrast to the correlation between the changes in global LUS score and the APP duration, the duration of APP was not correlated with the changes in oxygenation. This may be explained by the fact that prone-related oxygenation response is determined by the balance between the resolution of dorsal atelectasis and the formation of ventral atelectasis on the one hand, and the changes in lung perfusion, on the other hand [
22]. It is worth noting that in 35% of intubated patients in the study by Rossi et al., oxygenation decreased after prone positioning. They explained these negative changes by the prevalence of consolidated lung tissue, which was less likely to be recruited with prone positioning [
22]. Therefore, using the oxygenation response to assess the extent of dorsal recruitment during prone positioning might not be appropriate. Fossali et al. found similar results in a recent physiological study, where they evaluated 21 intubated patients with COVID-19 with CT scan and EIT [
23]. They found extensive alveolar recruitment in dorsal regions with prone positioning, along with alveolar derecruitment in ventral zones, which was, however, to a smaller extent than recruitment in the dorsal regions. No significant correlation between global or regional lung recruitment and oxygenation was found [
23]. Additionally, given that LUS can assess the severity of COVID-19 pneumonia as good as the chest CT scans in most symptomatic patients with confirmed COVID-19 [
24], our results support the utility of LUS to assess the lung morphology and monitor its changes with APP, especially if CT scan is not available or in patients at high risk for moving to the CT scan room. Moreover, we found that more patients in the treatment success group had improvement in dorsal zone LUS scores from 3 to 2 (consolidation to coalescent B lines) than in the treatment failure group. As we did not perform an extended LUS, we cannot exclude that this finding could be due to the existence of different pathologies (atelectasis vs pneumonia) [
25].
Prone positioning significantly improved oxygenation in the first pronation for both patients with treatment failure and success. However, oxygenation improvement associated with prone positioning persisted on the second and third days only in those patients with treatment success. Similar results were obtained in previous studies on non-intubated [
26] and intubated [
13] COVID-19 patients who received prone positioning. However, our previous study with non-intubated patients with COVID-19 found that patient oxygenation response to APP on the second day could predict the need for intubation [
13], while this was not significant in the current study. The discrepancy might be explained by the smaller sample size (71 vs 108) and the greater severity (average SpO
2/F
IO
2 of 100 vs 150 at the study enrollment). In contrast, patient response to the first APP session assessed by LUS was significantly associated with the treatment success, implying that LUS response to the first APP session was more sensitive than other physiological parameters such as oxygenation responses to APP in predicting the need for intubation.
This study has several strengths. First, to our knowledge, this is the first to report the predictive value of global LUS scores for intubation in COVID-19 patients receiving APP. Second, this is a two-center study with a large and predefined sample size to assess patient response to APP using LUS. Third, a blinded interpretation of the LUS was performed by two expert clinicians. Lastly, we assessed LUS scores in the first APP sessions on three consecutive days rather than a single session.
This study also has several limitations. First, the LUS assessment was performed only once daily, whether the APP response in the other sessions would have been different was unknown. Second, we did not evaluate other ultrasonographic variables such as diaphragm displacement or thickening fraction that were utilized to assess inspiratory effort during APP for COVID-19 patients [
27]. Third, the contribution of HFNC added to APP in the recruitment of the dorsal regions might be different from the combined use of noninvasive ventilation or conventional oxygen with APP, so our findings could not be generalized to patients treated with other noninvasive respiratory support therapies. Lastly, this study was designed to investigate the change in global LUS score after 3 days of APP in a single group, therefore, the results regarding predictors of intubation should be validated in another study that incorporates a control group of patients who will remain in the supine position, which might raise an ethical issue.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.