Patients, equipment, and study groups
All the LUS-guided recruitments, showing signs of lung recruitability (S-pattern development) performed between January 2019 and August 2021 were included in this retrospective study and constituted the Lung UltraSound Targeted Recruitment (LUSTR) group. The LRM was performed in the case of severe respiratory deterioration, expressed as S/F ratio below 200 [10]. The LMR was excluded from the study in case no S-pattern was detected during the procedure. LUS evaluation was executed by an expert neonatologist, experienced in LUS, using a high-frequency linear probe (9–16 MHz) with General Electric Medical System LOGIQ S8 or with Mindray DC8-EXP ultrasound machines.
The control group (Ox-group) was formed by rescue LRMs conducted with the standard oxygenation-guided procedure [11], as the personnel trained in LUS was not available. The oxygenation-guided LRMs were performed during the same period and selected to match those of the LUSTR-group, in terms of type of pulmonary pathology and severity of the disease (initial S/F ratio).
Infants enrolled were mechanically ventilated either with VN500 (Draeger, Lubecca, Germany) or with Sensor medics 3100A (Carefusion, San Diego, USA).
This retrospective study was approved by the local ethics committee (CEROM, n. 4510 Prot. 8317/2021 I.5/16), and conducted in accordance with the declaration of Helsinki.
All the infants were well sedated during the maneuver. No muscle relaxant was given.
Lung Recruitment Protocols
In the intervention group, LUS was systematically performed on all lung fields: anterior superior, lateral superior, posterior superior, anterior inferior, lateral inferior, and posterior inferior for the right and left sides [12].
If signs of lung consolidation or derecruitment (C-pattern with air bronchogram) were detected, after the exclusion of any modifiable causes of lung collapse (e.g., endotracheal tube misplacement), LUS-guided recruitment was initiated [7].
The neonate was positioned to maintain the most severely affected lung fields on the upper side, with constant ultrasound monitoring to visualize the progressive parenchymal reopening. LRM was performed through the stepwise increase of airway pressure, either positive end-expiratory pressure (PEEP) during conventional mechanical ventilation (CMV) or continuous distending pressure (CDP) during high-frequency oscillatory ventilation (HFOV).
Airway pressure was augmented by 1 cmH2O every 1–2 minutes. The recruitable lung was determined by the development of the S-pattern during pressure augmentation. The S-pattern is characterized by the presence of vertical hyperechoic artifacts starting from the reopening bronchi, representing a sonographic sign of lung recruitability. As these linear artifacts resemble the sun rays crossing the clouds, they were named Sunray(S)-lines. The pattern is characterized by the appearance of at least one S-line per field [7]. The pressure was increased stepwise until the S-pattern disappeared and converted into a B2 or B1 pattern, and the corresponding level of airway pressure was defined as opening PEEP/CDP (Fig. 1). The maneuver was interrupted before obtaining the S-pattern resolution in the case of clinical deterioration.
PEEP/CDP was reduced stepwise until LUS began to show signs of lung consolidation or aeration loss in the lung field previously monitored. The corresponding airway pressure was defined as closing PEEP/CDP (Fig. 1). The optimal PEEP/CDP was set right above the closing pressure. In the case of atelectasis development/persistence in other lung fields, the infant’s position was modified again to maintain the atelectatic lung field on the upper side and another LR was performed. A comprehensive LUS follow-up was performed 1 hour, 6 hours, and 12 hours after the end of the maneuver.
In the control group, the procedure was performed as described in previous reports [6, 11].
Data collection
The following clinical data were retrospectively retrieved from medical records: birth weight (BW); gestational age (GA); underlying lung pathology, defined as (i) respiratory distress syndrome (RDS) [13], resistant to surfactant administration, (ii) neonatal acute respiratory distress syndrome (nARDS), following neonatal sepsis [14], (iii) lung derecruitment during MV detected at chest X-ray or LUS; days of life at recruitment; incidence of bronchopulmonary dysplasia (BPD), defined as oxygen dependency at 36 weeks post-menstrual age in infants born before 32 weeks of GA [15]; length of stay; duration of mechanical ventilation; survival.
The following data were collected about LRMs: mode of ventilation; type of ventilator; FiO2, CDP/PEEP, SpO2, heart rate, and mean blood pressure at each time point (start, opening, and optimal pressure); duration of clinical/sonographic benefit, defined as hours without respiratory deterioration and/or echographic evidence of atelectasis; air leaks within 12 hours from LRM; the need for chest drain for air leaks within 12 hours from LRM. The presence of air-leak was suggested by clinical deterioration which prompted the X-ray execution in the Ox-group. While in the LUS group, this was excluded by the follow-up LUS examinations. To assess the severity of the lung disease we used the following parameters: the S/F ratio [10] and oxygen saturation index (OSI = CDP x FiO2/SpO2) [16].
In the LUSTR-group was collected the number of steps of pressure needed to obtain the S-pattern. Ultrasound images and videos were retrospectively analyzed by two independent operators to assess the presence of S-. No disagreements were reported in this phase.
Study bias
Selection bias was avoided including all the infants undergoing the rescue procedure (S/F) ratio below 200. The Ox-group was obtained by automatically matching for 2 relevant different variables. There was no follow-up, therefore no loss was experienced in this study. The multiple linear regressions were performed to correct for the most important confounding variables.
Primary and secondary outcomes
The primary outcome was the difference between the final and the initial S/F ratio (delta S/F ratio). A secondary efficacy outcome in patients undergoing HFOV ventilation was the difference between the final and the initial oxygen saturation index (delta OSI). A secondary safety outcome was the development of pneumothorax with and without the need for chest drain within 12 hours from the LR maneuver.
Statistical analysis
To compare the characteristics across the groups, Chi-square test of independence and an F-test were performed for categorical and continuous variables, respectively. A one-way nested ANOVA was conducted for each comparison group to determine whether the differences between the group means were statistically significant and to take into account the nested structure of the data as multiple observations are associated with the same patient. To this end, the response variable was log-transformed to normalize its distribution when needed.
A multiple linear regression model was used to estimate the association between the S/F ratio and relevant explanatory variables based on the theory. The same multiple linear regression model was used with OSI as the response variable. The mean was reported with the SD (standard deviation), and the IQR (interquartile range) with the median for the characteristics related to the observations and the patients, respectively. All model assumptions were checked for each model, furthermore, the models were evaluated for multicollinearity, without detecting any relevant effect. As a result of the multiple linear regression, we reported the unstandardized regression coefficient (B) which describes how much the dependent variable is expected to increase when an independent variable increases by one, holding all the other independent variables constant. We then reported the standard error for the unstandardized beta (SE B), in order to describe the variation around the estimates of the regression coefficient; the t-test statistic (t), and the p-value.
We performed an internal analysis of the LRs performed with the standard method and found a mean delta S/F ratio of 70+/-30 during the procedure. Assuming an improvement of 50% with LUS guidance, the minimal sample size to detect significance was 8 recruitments per group. However, to further avoid selection bias, we included all the LUS recruitment showing signs of lung recruitability during the study time. All the recruitments included in the LUSTR-group were matched through the software matching system with one control, based on the severity of the disease (initial S/F) and the type of lung disease. No missing data regarding the variables of interest (delta S/F and delta OSI) were reported. Variables with a significant number of missing data were excluded from the analysis.
All statistical analyses and matching were performed by a statistician (LA), using the statistical software R (Version 4.1.2, R Development Core Team, Vienna, Austria).