Mechanisms of oxygenation responses to proning and recruitment in COVID-19 pneumonia

We studied 25 COVID-19 pneumonia patients, admitted in the intensive care unit (ICU) of the Parma University, between March, 18th 2020, and January 29th 2021. The informed consent was obtained using a remote process after discharge. Protocol number of the ethical committee: Comitato Etico Azienda Ospedaliera-Universitaria di Parma, 779/2020/OSS/AOUPR. Although our intention was to consecutively study all admitted COVID-19 patients, this was impossible for logistic reasons (man power required to perform the study and availability of the dedicated CT scan). Within this frame, the patients were not arbitrarily selected, but studied whenever possible. Every patient was measured only at one single time point. Therefore, our analyses are not intended to provide longitudinal “follow-up” but to elucidate the physiological and CT characteristics of patients studied over that timeframe following their hospital admission (see supplementum for further details).

We chose 35 cmH₂O as the maximum airway pressure instead of 45 cmH₂O [5], considering the higher incidence of pneumothorax reported in COVID-19 compared to ARDS from other etiologies [6]. FiO₂ was maintained constant throughout the experimental steps.

By quantitative analysis of CT scan [7], we measured the lung anatomical variables as previously described [8]. In addition, comparing the CT scan in supine-5 and supine-35, we differentiated the atelectasis (“empty” and openable pulmonary units) from consolidation (“full” and non-openable pulmonary units).

Consolidation refers to a substitution of alveolar gases with material, while atelectasis refers to emptying the alveolar units from gases.

We quantified consolidated lung tissue as follows:

Therefore, we considered as consolidated, the fraction of non-aerated tissue which could not regain aeration at 35 cmH₂O of airway pressure, i.e., the amount of aerated tissue after the recruitment maneuver.

We quantified atelectatic lung tissue and atelectatic tissue fraction as follows:

To investigate the relationship between the non-aerated tissue fraction and venous admixture, we calculated the “apparent perfusion ratio” [9] which expresses the ratio between the perfusion of each gram of non-aerated tissue and the perfusion of each gram of aerated tissue (see supplement for derivation).where Qva/Q is the venous admixture according to Riley’s model [10].

A value of apparent perfusion ration = 1 indicates equal perfusion in non-aerated and aerated tissue, < 1 indicates relative hypoperfusion of the former, and > 1 its relative hyperperfusion.

Data are presented as mean ± standard deviation or median and interquartile range, as appropriate. The chi-square test or Fisher´s exact test of independence was used for categorical variables, T-Test for continuous variables and non-parametric test (Kruskal–Wallis test) in case of non-normal distribution. Linear regression was used to assess the relationship between continuous variables. A One-way ANOVA test for repeated measure was used to account for the repeated measures design and Tukey post-hoc test for multiple comparisons. Two-tailed p values < 0.05 were considered statistically significant. Rstudio for Statistical Computing with the Tidyverse package collection was used for analysis.

As shown in Table 1, the average time between hospital admission and study enrollment was 11 ± 6 days (from 3 to 24 days). At the time of study, the majority of the patients satisfied the criteria of moderate ARDS, according to Berlin criteria, and the overall hospital mortality was 32%. All patients were routinely treated with prone position.

The most relevant anatomical and physiological variables we measured in prone-5, supine-5 and supine-35 are shown in Table 2. As indicated, there were only few significant changes with repositioning from supine-5 to prone-5. In particular, the amounts of gas and the fractions of normally, poorly, and non-aerated tissue were similar for prone-5 and supine-5, while the atelectatic tissue fraction significantly decreased from 13 to 8%. Gas-exchange variables were similar between supine-5, and prone-5, while the respiratory system mechanics worsened from supine-5 to prone-5 due to an increase of respiratory system elastance, likely due to the increased stiffness of the anterior chest wall [11]. pH was slightly but significantly higher in prone-5 than in the other two conditions.

To reach an inspiratory airway pressure of 35 cmH₂O during the recruitment maneuver required the use of significantly larger tidal volume and driving pressures in supine-35, compared to supine-5 and prone-5. Therefore, the gas volume, the overinflated and the normally aerated tissue fractions were higher in supine-35, compared to supine-5 and prone-5, while the poorly and the non-aerated tissue fractions were significantly lower. The respiratory system elastance decreased significantly in supine-35, reflecting the reduction in non-aerated tissue. Regardless the overall improvement of tissue aeration, however, the gas-exchange variables did not improve and were similar to prone-5 and supine-5. In addition, the ventilatory ratio significantly deteriorated in supine-35.

The arterio-venous (a-v) O₂ content difference, and the central venous oxygen saturation, considered here as a surrogate of cardiac output, were similar in the three conditions we tested. However, during the supine-35, a slight but significant decrease of mean arterial pressure and heart rate was observed. Therefore, changes in cardiac output cannot be excluded. The apparent perfusion ratio, an indicator of the perfusion of non-aerated tissue relative to the perfusion of aerated tissue [9], almost doubled in supine-35, compared to supine-5 and prone-5.

Repositioning from supine-5 to prone-5 caused a significant change in lung shape. Indeed, in prone position the fraction of the total tissue mass in the upper half of non-dependent lung was 61.5 ± 0.05%, almost double that in supine position (32 ± 0.04%, p < 0.001). The tissue distribution is represented in Fig. 1, upper panels. As shown (panel C), in prone position the non-aerated tissue increased in the ventral levels of the lung while it decreased in the dorsal ones. In other words, in prone position new ventral atelectasis were formed, while the dorsal atelectasis present in supine position regained aeration and disappeared. The formation and dissolution of atelectasis was a function of the changes in hydrostatic superimposed pressure (see Fig. E1). Indeed, in prone position, the ventral atelectasis appeared because of the increased compression exerted by the weight of the lung above, while the dorsal atelectasis disappeared as the compression was released. As opposed to the non-aerated tissue, the well-aerated tissue increased in the dorsal regions during prone positioning and decreased in the ventral zones, compared to supine (Fig. 1, panel A).

Although the differences in PaO₂/FiO₂ and Q_VA/Q, between supine-5 and prone-5, did not reach statistical significance (see Table 2), their changes when going from supine to prone position were independently correlated both with the changes of the non-aerated tissues (i.e., the balance between “old” dorsal atelectasis, which disappeared, and “new” ventral atelectasis, which appeared) and with the changes of its apparent perfusion ratios (see Fig. E2 for the individual responses and, see Figs. E3 and E4 for regressions). It is worth noting that the consolidated tissue maintains its anatomical location (either dorsal or ventral) independently on the position (Fig. E5) and this carries three consequences: first, the dorsal consolidated tissue does not reopen and cannot contribute to improve oxygenation. Second, the ventral atelectasis anyway develops to the increased compressive forces (p = 0.012, Fig. E6). Last, the final result is a greater venous admixture in prone-5 than in supine-5 (p = 0.018, Fig. E7). Thirty-five percent of the patients experienced a negative PaO₂/FiO₂ difference in the prone position relative to the supine position, and 65% could be classified as non-responders, having undergone a PaO₂/FiO₂ increase ≤ 20 mmHg [12].

The tissue distributions along the sterno-vertebral axis in supine-5 and supine-35 are represented in Fig. 1, lower panels. As shown in panel D, the normally aerated tissue in supine-35 increased remarkably at each lung level compared to supine-5. The poorly aerated tissue (Panel E) was slightly more prevalent in supine-5 and decreased in supine-35, while the non-aerated tissue (Panel F) significantly decreased, primarily in the dorsal dependent regions. In the different levels from 1 to 10, the differences in the heights of the non-aerated tissue of supine-5 and 35 columns quantify the amount of atelectatic tissue which disappeared at 35 cmH₂O of inspiratory pressure (176.8 ± 187 g/patient). Although the differences in PaO₂/FiO₂ and Q_VA/Q, between supine-5 and supine-35, did not reach statistical significance (see Table 2), the changes of Q_VA/Q when increasing the airway pressure from 5 to 35 cmH₂O, were correlated with the changes of the non-aerated tissues (i.e., the re-aeration of the atelectasis) (see Fig. E8), while the perfusion ratio of consolidated tissue increased from 1.4 in supine-5, to 2.15 in supine-35 (p < 0.001). Of note, 48% of the patients experienced a decrease in PaO₂/FiO₂, during the recruitment maneuver (see Table E1 in supplement and Fig. E9 for the individual responses).

In Fig. 2 we illustrate the distribution of consolidated and atelectatic tissues in patients studied during the first-, the second- and the third week following their hospital admission. As shown, the consolidated tissue was lower in patients studied during the first week than in patients studied in the third week from hospital admission, and its fraction significantly correlated with the time elapsed between hospital admission and the study day (p = 0.013, Fig. E10). This increase in consolidation was paralleled by a similar increase of the respiratory system elastance (p = 0.03, Fig. E11). In contrast, the amount of atelectatic tissue did not change significantly with elapsed time.

The differences between anatomical, physiological and clinical variables recorded in the three different weeks are presented in Table 3. As shown, the PaO₂/FiO₂ variables were similar at entry to hospital and remained similar throughout the weeks. In contrast, the PaCO₂ and the ventilatory ratio were higher if the time elapsed from hospital admission to the day of study was greater (p = 0.008 and p = 0.01, Fig. E12).

Although only the consolidated tissue was significantly greater over the three weeks, there was a clear tendency to worsening of all the other anatomical variables explored. Of note, the frequency of patients who showed a negative delta PaO₂/FiO₂ to prone positioning and to recruitment increased progressively over time. Similarly, the PaCO₂ response to prone positioning (p = 0.047) and recruitment (p = 0.016), deteriorated significantly with the time elapsed from hospital admission to CT scan (Fig. E13).

Several phenomena occur contemporaneously when patients are shifted from supine to the prone position, as a more homogeneous distribution of tissue mass, gas-to-tissue ratio and an increase in chest wall elastance [13]. All these elements are of great relevance in providing a better distribution of stress and strain and, therefore, contributed to lung protection [14]. As far as oxygenation is concerned, however, the primary phenomenon is the balance between the release of dorsal atelectasis and its formation in the ventral zones. This density “redistribution” has been consistently observed in early ARDS [15, 16], where the tissue collapse in the ventral regions is associated with a near total re-expansion of collapsed dorsal lung units (see Fig. E14, panel A). However, if a significant amount of consolidated tissue is present in dorsal regions, as in our patients with COVID-19, the dorsal de-collapse in prone is very limited as the irreversibly consolidated tissue cannot reopen (see Fig. E14, panel B) while the ventral regions still bear the weight of the dorsal regions situated above them. This compression (i.e., the superimposed pressure) determines the formation of ventral atelectasis. Therefore, the lesser or greater amount of consolidated tissue within the non-aerated tissue in dorsal lung regions accounts for the huge variation of recruited/derecruited tissue between prone-5 and supine-5 positions, ranging from + 225 to -120 g (model in Fig. E15).

Although the prone position is largely used in COVID-19 to improve oxygenation, in our population we found that Delta PaO₂/FiO₂ was negative in 35% of the patients when in the prone position, and 65% could be classified as a “non-responder”, according to the previously used threshold of Delta PaO₂/FiO_{2 (prone-supine)} ≤ 20 mmHg [12]. We found two variables that were independently associated with the position-related PaO₂/FiO₂ changes, i.e., recruitment/derecruitment (the balance between dorsal and ventral atelectasis) and the changes of the apparent perfusion ratio. In experimental models [17‐19] and in normal subjects [20, 21], the perfusion does not change significantly with position, and the recruitment/derecruitment ratio fully accounts for the changes of PaO₂/FiO₂. In COVID-19 patients, however, where the loss of perfusion control is a characteristic trait [22‐25], we found that the changes of the perfusion ratio contribute to the changes of PaO₂/FiO₂. Indeed, an increase of perfusion ratio, i.e., a gravity dependent increase of perfusion of the ventral atelectasis dampens or reverses any PaO₂/FiO₂ rise due to recruitment.

Several studies reported recruitability and response to PEEP increase in COVID-19 patients, by using different techniques for recruitment assessment, from CT scan [26] to recruitment/inflation ratio [27]. Although the responses were variable, the majority of the studies observed a limited rectruitability and poor responses to PEEP [2, 27, 28]. The mechanism underlying these different responses, however, are still unclear.

In our population cohort, two phenomena appeared to occur simultaneously when the airway pressure was raised from supine-5 to supine-35. On the one hand, all the atelectasis present in supine-5 disappeared in supine-35 (as per our definition). For an unchanged perfusion of the previous atelectatic tissue, this phenomenon per se should promote an increase of PaO₂/FiO₂. On the other hand, perfusion diverted from previously atelectatic tissue to consolidated tissue, almost doubling its apparent perfusion ratio (see Table 3). These phenomena per se should lead to a decrease of PaO₂/FiO₂. Therefore, the changes of oxygenation during recruitment in these patients would depend by the balance between reopening of atelectasis and perfusion re-distribution.

Actually, the 12 patients who decreased their PaO₂/FiO₂ after increasing the pressure to 35 cmH₂O had less atelectatic tissue fraction to be recruited than the patients who increased their PaO₂/FiO₂ (0.19 ± 0.15 vs 0.45 ± 0.23; p = 0.002) (see Table E1 in supplement). In addition, the patients who decreased their PaO₂/FiO₂ had significantly higher consolidated tissue fraction (and increased perfusion at 35 cmH₂O), than did the patients in whom the PaO₂/FiO₂ increased by that maneuver (consolidated tissue fraction 0.27 ± 0.10 vs 0.18 ± 0.10, p = 0.04) (see Table E1 in supplement).

Our data do not allow us to prove the role of time on COVID-19 pneumonia evolution. Indeed, our patients were not studied longitudinally. We observed, however, that the frequency of higher elastance, lower gas-volume, higher PaCO₂ and ventilatory ratio, greater consolidation and lack of response to prone positioning and lung recruitment were higher in the patients studied in the third week rather than those studied in an earlier stage of the disease. As the prevalence of these changes are coherent with a progressive evolution of the lung pathology towards organizing pneumonia and “fibrosis-like” status [29‐32] it is tempting to speculate that time had a key-role on the evolution of COVID-19 pneumonia. A decreased response with time to prone position and recruitment manoeuvers, however, has been previously described in COVID-19 and in non-COVID-19 ARDS [31, 33, 34].

A major limitation of the study is the relatively small sample size of the study cohort and lack of longitudinal design, as different patients were studied at different time-points. Another limitation is the lack of formal randomization of the prone-supine positions sequence. These limitations were largely due to the complexity of the protocol and to logistical difficulties imposed by the pandemic. Indeed, this study was feasible only on days when sufficient workforce and COVID-19 dedicated CT were available. We should note, however, that we are dealing with a disease (similar in all patients, and not with a syndrome). Moreover, most prior advances in understanding the pathophysiology of ARDS by CT scan, such as the baby lung [35], density redistribution in prone position [15] and mechanisms of opening pressures [36‐38] were achieved by carefully studying a similar or lower number of patients. In this study we used a total recruitment pressure of 35 cm H₂O, compared with the standard value of 45 cm H₂O. Despite our choice could lead to an overestimation of the consolidated tissue, we opted for this choice for safety reasons. Another shortcoming of this study is the lack of full hemodynamic data, which impedes more thorough understanding of gas-exchange variations. In addition, the Qva/Q was measured using a central venous instead of a mixed venous blood.

In conclusion, we show that in unresolving COVID-19 pneumonia, the respiratory mechanics and the gas-exchange response to prone positioning and recruitment largely depend on the following two factors: perfusion dysregulation and the amount of consolidated tissue. As the amount of consolidated tissue was different among patients studied at different weeks since admission, it is possible that the respiratory treatment administered should be changed according to the stage of the disease.