Investigation of high flow nasal cannule efficiency with electric impedance tomography based parameters in COVID-19 adults patients: a retrospective study

Patient selection

This study was planned after approval from the University Ethics Committee (approval no: 2022/05-01) as a retrospective review of the hospital data of 20 concecutive patients who were diagnosed with COVID-19 confirmed by polymerase chain reaction (PCR) test, followed in the COVID ICU, and diagnosed with ARDS according to the Berlin criteria (Table 1) (Ferguson et al., 2012).The informed written consent form was obtained from the patients. Posterior-anterior chest X-rays and thorax computed tomography images of the patients were consistent with the typical infiltration and ground glass images of COVID-19 ARDS.

Exclusion criteria from the study were accepted as being younger than 18 years of age, being pregnant and lactating, having a body mass index (BMI) above 50 kg/m^2, having a rib cage malformation, presence of pneumothorax, and conditions where EIT monitoring is contraindicated (patients with automatic implantable cardioversion defibrillator, automatic drug pumps, patients with chest skin injury (Kunst et al., 1999).

The patients, which followed for COVID-19 ARDS and administered respiratory support with HFNC, demographic data, body mass indexes (BMI), APACHE-II score (acute physiology and chronic health evaluation), ratios of PaO₂ (partial oxygen pressure) in blood gas at baseline and during treatment, to FiO₂ (percentage of oxygen delivered in HFNC) (PaO₂/FiO₂ ratios), respiratory rate (RR), peripheral oxygen saturation (SpO₂), ROX index (SpO₂/FiO₂ ratio to RR) and hemodynamic parameters, including heart rate (HR) and mean arterial pressure (MAP), position information during therapy (supine-prone), flow values applied in HFNC, and EIT images were obtained retrospectively from the recordings. All of the blood gases taken from the patients were taken at least 20 min after the treatment in order to reflect the position and flow changes in the blood gas parameters. In addition, if intubation was performed, the days of intubation, the days of discharge from ICU, and mortality rates were retrospectively scanned from the data registration form.

All the patients in our research group were given all medical treatments (Favipiravir, corticosteroids, anticoagulant and if necessary, anabiotic etc.) in accordance with the National Covid Treatment Guideline published by the Ministry of Health of the Republic of Turkey (Kurulu, 2021).

Study protocol

In the study protocol, the flow rate of the HFNC was increased to two preset levels (30 L/min and 50 L/min), and each flow level was maintained for 20 min to a minimum. The reason for holding a minimum of 20 min in each position is that this is the minimum time required for any changes to be reflected in blood gas measurements. As a part of routine care in our clinic, arterial blood gas measurement is performed 20 min after each change. Because it is common practice to wait 20 to 30 min after the FiO₂ is changed before arterial blood gas measurements are obtained, 10 min may be adequate unless the patient has obstructive lung disease which requires a longer equilibration time (Hess et al., 2016). For this retrospective study, data of patients admitted to ICU with respiratory failure due to COVID-19 were obtained from the hospital data system and records on the EIT. In our clinic, this patient group is followed in two different flow rates (30 L/min and 50 L/min) and two different positions as part of routine care. Patients undergo lung imaging with EIT during HFNC treatment. In order to find the optimum flow rate during the treatment, arterial blood gas is taken at least 20 min after the flow level and position change. We think that it should take 20–30 min for the change to be reflected in the arterial blood gas.

Oxygen levels (FiO₂) in the HFNC were adjusted so that the saturation levels of the patients would not decrease from 90–92% (for patients with Type 2 respiratory failure, the saturation would not fall below 88–90%).

As invasive mechanical ventilation indications: include level of consciousness (Glasgow coma score <12), cardiac arrest/arrhythmias, severe hemodynamic instability (norepinephrine>0.1 µg/kg/min and persistent or worsening respiratory status and lack of oxygenation (despite HFNC flow ≥ 50 L/min and FiO₂>100; PaO₂<60 mmHg), respiratory acidosis (PaCO₂>50 mmHg, pH <7.25), respiratory rate >30 bpm or inability to clear secretions, and a ROX index below 2.85 were accepted (Mauri et al., 2017).

For all patients followed up, the temperature was set to 37 °C with the HFNC (Optiflow, Fisher & Paykel Healthcare, Auckland, New Zealand) humidifier (MR850, Fisher & Paykel Healthcare, Auckland, New Zealand), and oxygen was administered nasally with a medium silicone nasal cannula (RT050/051, Fisher & Paykel Healthcare, Auckland, New Zealand). Heat and humidified high flow nasal cannula or as most call it, hi flow nasal cannula (HFNC), is not just a standard nasal cannula cranked up to very high flow rates. It actually takes gas and can heat it to 37 °C with a 100% relative humidity and can deliver 0.21–1.00% fi02 at flow rates of up to 60 liters/min. The flow rate and FiO2 can be independently titrated based on the patient’s flow and fi02 requirements (Sharma et al., 2023). The patients were asked to breathe through the nose, and their mouths and noses were covered with a simple surgical mask due to the risk of contamination. The different flow rates and position information in the study were recorded as follows:

T0: Baseline measurements are taken in the supine position before any therapy is started.

T1: HFNC in supine position with a flow rate of 30 L/min.

T2: HFNC in supine position with a flow rate of 50 L/min.

T3: HFNC in prone position with a flow rate of 30 L/min.

T4: HFNC in prone position with a flow rate of 50 L/min.

In routine practice in our clinic, all patients admitted to the ICU are primarily treated in the supine position. However, in diseases such as C-ARDS where oxygenation does not improve in the supine position, patients are placed in the prone position to correct ventilation-perfusion mismatch. Therefore, the patients admitted to the ICU were first placed in the supine position and then in the prone position. For this reason, our research also accepted the supine position as the baseline.

EIT measurements

EIT measurements were performed on the Pulmovista 500 device (Dräger Medical, Lübeck, Germany), which is part of routine care in our clinic. In our clinic, instead of lung imaging with CT, imaging with EIT is applied to patients followed up with C-ARDS. For application, the silicone EIT belt with 16 surface electrodes was placed around the patients’ rib cage in the fourth intercostal space and then connected to the EIT monitor for visualisation. The reflections from tissues were measured by applying alternative current to lungs and surrounding tissue. The stimulation frequency and amplitude were adjusted automatically by the EIT device. EIT measurements were performed continuously at 20 Hz (Tomasino et al., 2020). When the patients were turned to the prone position, attention was paid to tying the EIT belt in the same spots by placing marker points so that the position of the EIT belt would not change. In addition, the data is digitally filtered with a cut-off frequency of 0.67 Hz to avoid reflection of heart-related impedance changes (Tomasino et al., 2020). EIT scans consist of 32 × 32 colour-coded matrix impedance images (Tomasino et al., 2020).

EIT data

In the data obtained with EIT, lung regions are divided into four regions of interest (ROI): by recording regional changes of waveforms resulting from ventilation, non-dependant ROI1 and ROI2 show right and left ventral lung regions; ROI3 and ROI 4 show dependent right and left dorsal lung areas (Zhang et al., 2020). In the supine position, both lung portions of the patient in the ventral region are defined as non-dependant, and both lung portions in the dorsal region are defined as dependent regions. While ventilation is high in non-dependant lung regions, perfusion is high in dorsal lung areas defined as dependant. It is defined as ventilation-perfusion mismatch, especially in ARDS patients. Similar to this definition, ROI areas showing ventral lung areas are defined as non-dependant, while ROI areas showing dorsal lung areas are defined as dependant. The total ventilation of the four ROI areas is 100% and approximately 25% in each ROI regions (Bickenbach et al., 2017). The ROI ratio (also called ROI ratio or impedance ratio) divides the ventilation activity of the ventral region by the dorsal ventilation activity of the fEIT images, and may indicate both lung recruitment and derecruitment (Tomasino et al., 2020). An ROI ratio close to 1 was accepted as indicating a homogeneous distribution of ventilation (Zhao et al., 2009).

Dräger PulmoVista 500 device is used in our clinic. In the Pulmovista 500 device, the lung is divided into 4 ROI (horizontal or quadrants) areas. A region of interest (ROI) is a user-defined area within a status image. The image can be divided into four ROIs and arranged either horizontally, as quadrants or in a customised way. The area covered by each ROI is represented by the corresponding regional impedance waveform and the regional numeric value (Teschner, Imhoff & Leonhardt, 2015).

ROI ratio is the ratio between the mean values of ventral (ROI 1 and ROI 2) and dorsal areas (ROI 3 and ROI 4) and represents a reliable measures of the ventilation distribution (Tomasino et al., 2020). Although the ROI ratio value close to 1 is accepted as an indicator of homogeneity, in another recent study, this ratio was accepted to be between 0.6 and 1.5 (Tomasino et al., 2020; Yang et al., 2021). However, in our study, we took the ROI ratio as 1 as a reference when evaluating the data.

The present study examined the proportional changes of ROI areas in different flows and positions, and the homogeneity of ventilation was determined. Image reconstruction was performed using the algorithm in the software EIT Data Review Tool (Dräger Medical, Lübeck, Germany). EIT data were analysed offline using MATLAB R2015 (The MathWorks, Inc., Natick, MA, USA) as programmed software.

EIT data analysis

To reduce the heterogeneity of spontaneous breathing when analysing EIT data, an analysis of 5-minute continuous EIT data (data averaged) was performed at five time points (T0, T1, T2, T3, T4) (Kunst et al., 1999). Three parameters were examined to analyse the EIT data of patients treated with HFNC in spontaneous respiration using the software program. In addition to the software program used, it is possible to obtain the relevant parameters with the help of the formulas given below.

The GI index was introduced to quantify tidal volum distribution within the lungs and aims to summarize the complex pulmonary impedance distribution pattern using a single numeric value (Eq. (1)) (Putensen et al., 2019). The ideal value for healthy individuals is 0.5 (Bickenbach et al., 2017). Increased GI index is in parallel with lung injury (Zhao et al., 2009). It has been used to set the optimum positive end-expiratory pressure (PEEP) value in studies (Putensen et al., 2019). The lowest possible GI index corresponds to the optimum PEEP level (Putensen et al., 2019). (1)${\displaystyle GI=\frac{\sum _{X{Y}_{\in \mathrm{lung}}}|{\mathrm{DI}}_{XY}-\mathrm{Median}\left({\mathrm{DI}}_{\mathrm{lung}}\right)|}{\sum _{X{Y}_{\in \mathrm{lung}}}{\mathrm{DI}}_{XY}}}$

DI = differential impedance, DI_xy = a defined pixel lung region, and DI_lung = all pixels represented in lungs (Zhao et al., 2009).

The regional ventilation delay index (RVD index) was developed to further analyse regional ventilation (Eq. (2)) (Muders et al., 2012; Wrigge et al., 2008). It expresses the time and delays from the beginning of inspiration to the opening of the closed alveoli (Putensen et al., 2019). Although the RVD index was initially developed for slow flow manoeuvres during invasive mechanical ventilation, it was later used for spontaneous breathing studies (Bickenbach et al., 2017). To avoid confusion, the RVD in spontaneous breathing studies was named RVD_sponbreath (Bickenbach et al., 2017). The value presented in our study is the RVD _{sponbreathvalue}. (2)${\displaystyle RV{D}_{\dot{I}}=\frac{t_{i,40\text{\%}}}{T_{inspiration,global}}\times 100\text{\%}}$ t_i40% indicates the time required to reach 40% of the maximum inspiratory impedance change, and T_{inspiration,global} indicates the expiration time calculated from the global impedance curve (Muders et al., 2012).

Center of ventilation (CoV) is a measure that defines the spatial distribution of ventilation (Putensen et al., 2019). It was defined as the weighted average of the geometric centres of lungs ventilation in the dorsal-ventral and right-left direction (eqn-(3)) (Frerichs et al., 1998). A CoV value of 50% indicates that the ventilation distribution is centred in the dorsal-ventral direction of the thorax (Putensen et al., 2019). A decrease in CoV values indicates that ventilation shifts towards non-dependent ventral lung regions due to the collapse of alveoli in dependent dorsal lung regions (Putensen et al., 2019). (3)${\displaystyle CoV=\frac{\sum \left(y_i\times T{V}_i\right)}{\sum T{V}_i}\times 100}$ TV _i represents the impedance change in EIT images, and the other value, Y_i, is the pixel height scaled dorsal and ventrally (Frerichs et al., 1998).

As a result of pairwise comparison of repeated measurements of EIT parameters, respiratory and hemodynamic parameters

Table 3 shows a significant difference between repeated measurements of the measurement parameter in terms of GI index (p < 0.05). Among the five measurements made for the GI index, It was observed that the difference between the mean measurements taken at T0–T1, T0–T2, T0–T3, T0–T4, T1–T3, T1–T4, T2–T3, T2–T4 and T3-T4 time intervals was statistically significant (p < 0.05). On the other hand, GI index measurement at T1–T2 time points was not found to be significant (p = 0.06)(Table 3; pages 12, 13,14).

Table 3 shows a significant difference between repeated measurements of the measurement parameter in terms of CoV (p < 0.05). Among the five measurements made for CoV, the difference between the mean measurements taken at T0–T3, T1–T3, T2–T3 and T3–T4 time intervals was statistically significant (p < 0.05). However CoV measurements at T0 and T1, T2, and T4 time points were not significant (p = 0.99). In addition, CoV measurements at T1 and T2–T4 time points and T2–T4 time points were not found significant (p = 0.99). (Table 3; pages 12, 13,14)

Table 3 shows a significant difference between repeated measurements of the RVD index measurement parameter (p < 0.05). Among the five measurements made for RDV, the difference between the mean measurements taken at T2–T4 time intervals was found to be statistically significant (Table 3; pages 12, 13,14) (p < 0.05).

Table 3 shows a significant difference between repeated measurements of the ROI ratio measurement parameter (p < 0.05). Among the five measurements made for ROI ratio, it was observed that the difference between the measurement averages taken at each time interval was statistically significant (Table 3; pages 12, 13, 14)) (p < 0.05).

Table 3 shows a significant difference between repeated measurements of the P/F ratio measurement parameter (p < 0.05). Among the five measurements made for the P/F ratio, it was seen that the difference between the measurement averages taken at all time intervals was statistically significant (Table 3; pages 12, 13, 14) (p < 0.05).

Table 3 shows a significant difference between repeated measurements of the RR measurement parameter (p < 0.05). It was observed that the difference between the means of measurements taken at all time intervals from the five measurements made for RR was statistically significant (Table 3; pages 12, 13, 14) (p < 0.05).

Table 3 shows a significant difference between repeated measurements of the MAP measurement parameter (p < 0.05). Among the five measurements for MAP, the difference between the mean measurements taken at T0-T3 time intervals was found to be statistically significant (p < 0.05). In addition, there was no significant difference in MAP measurements between T0 time points and T1, T2, T4 time points, T1 time points and T2, T3, T4 time points, and T2 and T3, T4 time points (Table 3; pages 12, 13, 14).

The change in evolution of estimated marginal means of GI index, CoV, RVD_{sponbreathvalue}, P/F ratio, respiratory rate, mean arterial pressure and ROI ratio are provided in Figs. 1–8. EIT images is reconstruction by MATLAB in Fig. 9.

Limitation

Our study has all the limitations of a retrospective study. Also, limited equipment and personnel were employed during the pandemic. In our study, in which the effects of HFNC application with EIT have monitored in patients followed up with the diagnosis of C-ARDS, the sample size was limited. However, the sample size could not be calculated since there has been no previous study on this subject in the literature. It is necessary to carry out studies with larger sample groups in the future. Further studies are needed in the future for the effect of EIT in visualizing ventilation.

Conclusion

The COVID-19 pandemic still remains a cause for global concern. EIT is a non-invasive, radiation-free clinical imaging tool. EIT has the potential to monitor the effectiveness of the prone position and to determine flow rates in HFNC treatment in COVID-19 ARDS patients. Positioning the patient, which can simply be summarised as “keep the healthy lung down”, is an effective treatment option to improve ventilation. According to the authors’ literature knowledge, our study is promising because it is the first study in the world to examine HFNC treatment in COVID-19 patients with EIT and the first EIT study in our country.