Progressive changes in pulmonary gas exchange during invasive respiratory support for COVID‐19 associated acute respiratory failure: A retrospective study of the association with 90‐day mortality

Ratio of arterial pressure of oxygen and fraction of inspired oxygen (P/F ratio) together with the fractional dead space (Vd/Vt) provides a global assessment of pulmonary gas exchange. The aim of this study was to assess the potential value of these variables to prognosticate 90‐day survival in patients with COVID‐19 associated ARDS admitted to the Intensive Care Unit (ICU) for invasive ventilatory support.


Editorial Comment
In this single-center study of COVID-19 patients, assessment of the oxygenation and fractional dead space utilising blood gases and measurements of expired carbon dioxide during the first week of ICU admission was used to predict 90-day mortality.Whilst not useful to predict outcomes from individual patients, there was a clear association between both parameters and 90-day mortality, stronger predictive value compared to standard ICU mortality assessment tools.This suggests that assessment of these markers could be helpful to assess outcome if respiratory failure is the dominant cause of ICU admission.

| INTRODUCTION
[4] The fractional dead space, the proportion of dead space volume (V d ) to the tidal volume (V t ), V d /V t , represents the volume of ventilated air that does not actively participate in gas exchange. 5Increased V d /V t in the early phase of ARDS is also associated with an increased risk of death. 6e P/F ratio together with the V d /V t provide a global assessment of pulmonary gas exchange.The potential value of these combined variables to prognosticate the outcome in patients with COVID-19 associated ARDS admitted to the Intensive Care Unit (ICU) formed the impetus for this study.By including patients undergoing invasive mechanical ventilation in ICU, we were able to collect granular data for pulmonary gas exchange during the critical phase of the disease beginning from admission to ICU.This is an important novel aspect of this study since most previous reports have only used average values for data captured with variable frequency over several days in the initial phase of COVID-19 associated ARDS.
The aim of this study was to describe the progression of P/F ratio and V d /V t over 4 weeks after admission to ICU.We specifically aimed to assess the predictive value of both P/F ratio and V d /V t for 90-day mortality in patients treated with invasive respiratory support for COVID-19 associated ARDS.

| Data collection
Patient characteristics included age, gender, body mass index (BMI), medical comorbidities (hypertension, cardiovascular disease, diabetes, chronic obstructive respiratory disease (COPD) and/or asthma, renal disease, smoking history, and immunosuppressive drug therapy).
Treatment data included length of stay (LOS) in ICU (LOS-ICU), use of prone positioning, anticoagulation treatment during the first week and steroid treatment in the ICU.
Gas exchange data were based on arterial blood gases collected during ICU admission with concurrent volumetric end-tidal capnography (etCO 2 , Microstream© Medtronic, USA).Patient characteristics, treatment, and gas exchange data were retrospectively collected from the electronic medical records.

| Mechanical ventilation, prone position, and treatment with steroids
Patients were ventilated in a pressure regulated volume control mode with tidal volumes of 6-8 mL/kg, positive end-expiratory pressure (PEEP) between 12 and 17 cmH 2 O and peak inspiratory pressure < 40 cmH 2 O according to institutional protocol.
Patients were considered for prone positioning at the discretion of the physician-in-charge if an F i O 2 > 0.7 was persistently needed to maintain a P a O 2 > 80 mmHg.The prone position was maintained for at least 16 h, 7 unless signs of intolerance developed.Patients admitted after July 2020 were commenced on a standard steroid regimen of 6 mg intravenous betamethasone once daily for 10 days 8 as treatment with steroids became routine practice during the pandemic.

| Arterial blood gases
Arterial blood gas (ABG) samples were collected at the discretion of the ICU-nurse and/or physician-in-charge.The F i O 2 and etCO 2 values were recorded at the time of every ABG and registered in the blood gas analyzer (Rapid Point 500© Siemens Healthineers, Germany) connected to the electronic medical record.Blood gases with missing values were excluded during data collection.All P a O 2 , P a CO 2 , and etCO 2 values in kPa were converted to mmHg.
The P/F ratio was calculated as: The V d /V t was calculated using Bohr's equation modified by Enghoff:

| Outcomes
The primary outcome was overall survival censored at 90 days following admission to ICU.Patient characteristics are reported in Table 1.The median age was 64 years with a majority of males (75%).The median BMI at ICU admission was 29 kg/m 2 and the most common comorbidity was hypertension (46%).A majority of patients were treated with prone position for 16 h at least once.The ICU-mortality was 23% (n = 30) and the 90-day mortality was 30% (n = 38).Patients who were dead at 90 days were older and they were ventilated and stayed in ICU for longer.

| Blood gases
In total, 19,396 arterial blood gas (ABG) samples were analysed with an average of nine ABGs collected over 24 h in each patient.The 24 h mean P/F ratio and mean V d /V t were calculated from daily data for all patients treated with invasive mechanical ventilation up to 4 weeks in ICU. Figure 1 shows the progression of mean P/F ratio over the first 28 ICU days.3).The ROC AUC's for the composite marker were consistently higher than the ROC AUC for the combination of SAPS II and age on admission, AUC 0.71 (95% CI 0.61; 0.81) (see Figure S1).
The results of the Cox regression analysis and HR for each parameter are shown in Table 3.For every 10 unit increase in the mean P/Fratio during the first 0-6 days in ICU, the HR for death within 90 days after ICU-admission was 0.85 (95% CI 0.77; 0.94), meaning that the risk for death at day 90 decreased by 15%.For every 10 mmHg increase in the mean P/F ratio during days 7-13 in ICU, the HR for death at day 90 was 0.82 (95% CI 0.75; 0.89).S1).
The overall 90-day mortality was 30% in the whole patient cohort.An increasing 90-day mortality was noted in patients with lower P/F ratio than 160 during the first week in ICU, and in patients with higher V d /V t than 0.25 during the same period (see Figure 4).

| DISCUSSION
This study analyzed a granular dataset of almost 20,000 measurements of pulmonary gas exchange performed in 130 patients with COVID-19 associated ARDS admitted to ICU for invasive mechanical ventilation.The overall 90-day mortality was 30% and progressive changes in mean P/F ratio and V d /V t were both associated with mortality.For every 10 units increase in mean P/F ratio over the first and second weeks in ICU, the risk of death at day 90 was reduced by about 20%.Conversely, for every 10% increase in mean fractional dead space, V d /V t , the risk of death at day 90 increased by about 60% using data for the first ICU week and close to doubled using data for the second week in ICU.The decrease of P/F ratio and rise in V d /V t during the first 14 ICU days were associated with higher risk for 90-day mortality with respective cut-offs at 160 and 0.25, respectively, for the first week, and 144 and 0.26 for the second week.
A mean P/F ratio below 160 in the first week following admission to ICU was associated with a clear separation in 90-day survival and hence of prognostic value.2][13] Palanidurai et al. re-analyzed the data of seven ARDS network trials of non-COVID-19 ARDS patients and showed that the PEEP to P a O 2 /F i O 2 ratio was superior to the P/F ratio for the prediction of in-hospital mortality in ARDS. 14It is known that PEEP affects P a O 2 and P/F ratios as well as their predictive validity for mortality in ARDS and classification of ARDS.The P/F ratio at lower PEEP settings result in more patients being labelled as having severe ARDS and higher PEEP settings result in patients labelled as having mild or even no ARDS. 15In our study almost all patients were ventilated with a PEEP greater than 12 cmH 2 O, which theoretically might reduce the number of false positive severe ARDS.
Whilst full intensive care treatment and measures are often required for the first week to stabilize patients with COVID-19 ARDS, prognostication in the second week and onwards is valuable.Our findings related to mean P/F ratio may be helpful in this regard to guide resource planning and admission logistics in a pandemic setting, particularly when demand for healthcare becomes overwhelming.It may also guide family conversations to consider visiting privileges in the context of social restrictions and to promote more accurate expectations.A mean P/F ratio <140 during the second week of mechanical ventilation, consistent with moderate ARDS, was prognostically unfavorable in this study with clear separation between the groups of survivors and non-survivors at 90 days.This was also evident from the Kaplan-Meier analysis.This highlights the importance of mean P/F ratio progression even within a moderate disease stratum.After 3 weeks the confidence intervals were quite wide, reflecting the low number of patients remaining in ICU in our study cohort.
Furthermore, mortality at this stage of the ICU admission would likely be multifactorial and include non-COVID related morbidity.
The impact on mortality by increased V d /V t in patients with ARDS was first described in 2002 and later confirmed by other reports. 6,16,17ese studies used Enghoff's modification of the Bohr equation.9][20][21] There are several methods for quantifying the fraction of dead space ventilation in mechanically ventilated patients, and any one single measure is unlikely to convey a complete picture in all clinical settings.Generally, the dead space fraction is considered to be normal if below 0.3. 22 this study, a progressive rise in V d /V t was associated with a higher risk of 90-day mortality during the first and second week, whilst mean values remained within the normal range.In contrast, other studies investigating the early course of non-COVID and  microvessels, with a prevalence twice higher than critical non-COVID-19 patients. 23In our cohort we could not prove the extension of thrombotic pulmonary complications as regular chest CT pulmonary angiographies were not performed.Whilst the levels of fibrinogen and D-dimers were followed regularly, the correlations between those potential markers of increased thrombotic pulmonary complications and V d /V t were not further explored since it was beyond the scope of this study.7][28][29][30][31][32] In contrast to our study, most of these studies used data gathered during the first 24 h in the ICU, which as previously mentioned may not be entirely representative of the state of respiratory failure in these patients.Thus, more studies with a wider range of ARDS patients are needed to attempt the creation of a composite marker for the prognostication of mortality, but we still believe that serial measurements of respiratory parameters such as P/F ratio and

FUNDING INFORMATION
Funding has been received from University of Gothenburg, Sweden.
This single-centre, retrospective observational cohort study was conducted at the Department of Anaesthesiology and Intensive care at Sahlgrenska University Hospital/Mölndal.Ethical approval for the study was granted by the Swedish Ethical Review Authority Sweden (approval 2021-04760 and 2024-00663-02).Patients with acute respiratory failure due to COVID-19 were admitted to our ICU from 24 March 2020 to 20 May 2021.This time period represents the time between the first pandemic ICU admission and the time when admissions primarily for COVID respiratory failure became so infrequent that a pragmatic decision to cease inclusion was made.Study inclusion criteria were: (1) age ≥ 18 years, (2) COVID-19 pneumonia confirmed by a positive SARS-CoV-2 PCR nasopharyngeal test, (3) acute respiratory failure needing mechanical ventilatory support during the stay in ICU, and (4) admission to ICU without limitations to medical therapy.The exclusion criteria were: (1) patients were infection free within 7 days of ICU admission (2) acute respiratory failure needing only oxygen therapy via high flow nasal cannula or non-invasive ventilatory support; (3) death deemed imminent within 24 h of ICU admission.

3 | RESULTS 3 . 1 |
Patient characteristics A total of 154 patients were admitted to the ICU with COVID-19 pneumonia during the study period.Following exclusion of 14 patients who were infection free within 7 days of admission to the ICU, nine patients who did not require invasive ventilatory support during the stay in ICU and one patient who died within 24 h, 130 patients were included in the study.Intubation and initiation of mechanical ventilation was performed in other ICUs in 117 patients who were subsequently transferred to the study ICU.Five patients were intubated and treated beyond 2 days in other ICUs prior to transfer.Two patients were transferred out of the study ICU whilst still intubated.

F I G U R E 1
The progression of mean P/F ratio over the first 28 ICU days.Mean values for patients alive at 90 days are represented by the blue line and 95% confidence intervals by the blue shaded area.Mean values for patients dead at 90 days are represented by the red line and 95% confidence intervals by the red shaded area.The numbers of patients at risk in both groups are listed above the time axis.

at 1 .
97 (95% CI 1.42; 2.73), meaning that the risk of death almost doubled.The P/F ratio and V d /V t remained significant in the Cox regression sensitivity analysis using imputated data (see Table

COVID- 19
associated ARDS have shown that the mean V d /V t was markedly elevated with greater fractional dead space in non-survivors compared to survivors.6,17The aetiology of high physiologic dead space in ARDS, including cases triggered by COVID-19 infection, may be explained by several factors including permissive hypercapnia as part of protective ventilation strategy with low tidal volumes, and/or diffuse pulmonary endothelial injury associated with macro-and microthrombi formation.COVID-19 patients frequently develop a pro-coagulative state caused by virus-induced endothelial dysfunction, cytokine and complement cascade hyperactivation, with important prognostic implications.Lungs are the most affected organs and critically ill patients often show thrombotic lesions in pulmonary

F I G U R E 4 A
Kaplan-Meier survival analysis of 90-day survival for mean P/F ratio (left) and mean V d /V t (right) cut-off values from the first ICU-week.Left: The blue line represents patients with mean P/F ratios below 160 mmHg and the red line patients with mean P/F ratios above 160 mmHg during the first ICU-week.Right: The blue line represents patients with mean V d /V t below 0.25 and the red line patients with mean V d /V t above 0.25 during the first ICU-week.
added value to our understanding of both the progression and prognosis of COVID-19 associated ARDS.This study is limited by its single centre, retrospective, observational design.Since ventilator settings were not routinely recorded in a digital format, detailed data on mechanical ventilation could not be effectively retrieved for the purpose of this study.Such data would have allowed us to assess the potential effects of ventilator-induced lung injury (VILI) and to study the impact of PEEP on P/F ratio and V d / V t .Not having access to detailed data on mechanical ventilation is thus a limitation that warrants a more cautious interpretation of the study results.Changes to the treatment of COVID-19 associated ARDS occurred during the study period, such as increased dose of anticoagulation and introduction of betamethasone after June 2020.These factors overall improved the outcome for COVID-19 patients and could theoretically attenuate the ability of P/F ratio and V d /V t to predict patient outcomes.5| CONCLUSIONSThe trajectory of mean daily and weekly P/F ratios and V d /V t were important predictors for 90-day mortality in patients with COVID-19 associated ARDS.The mean P/F ratio was within an ARDS range from the start of invasive ventilation, whilst the V d /V t values remained within normal ranges initially with a subsequent deterioration in nonsurvivors.The magnitude and progression of impaired gas exchange separated 90-day survivors from non-survivors after adjustment for age and SAPS II.Further multicenter studies are needed to confirm the predictive value of mean P/F ratio and V d /V t in COVID-19 associated ARDS patients, with the results of this study supporting their utility in any future pandemic infectious respiratory failure disease.AUTHOR CONTRIBUTIONSYlva Konsberg: Conceptualization (equal), data curation (lead), formal analysis (equal), methodology (equal), project administration (equal), visualization (equal), writing -original draft preparation (lead), writingreview and editing (equal).Anders Åneman: Conceptualization (equal), formal analysis (equal), methodology (equal), software (equal), writing -review and editing (equal).Fredrik Olsen: Formal analysis (equal), writing -original draft preparation (equal), writing -review and editing (equal).Fredrik Hessulf: writing -original draft preparation (equal).Bengt Nellgård: Funding acquisition (lead), resources (lead), supervision (equal), writing -review and editing (equal).Mathias Hård af Segerstad: Conceptualization (equal).Keti Dalla: Conceptualization (equal), formal analysis (equal), methodology (equal), project administration (lead), supervision (lead), writing -review and editing (equal).
Patient characteristics and clinical data in the ICU for all patients and subgroups, alive at day 90 or dead at day 90.
The HR for every 0.1 unit (10%) increase in the mean V d /V t during the first week in ICU was 1.61 (95% CI 1.20; 2.16), meaning that the risk for death at day 90 increased by 61%.During days 7-13 of ICU admission, the HR for death at day 90 for every increase in V d /V t by 0.1 increased further to T A B L E 1 LOS-ICU, median (IQR) 18 (11-27) 16 (10-24) 24 (18-30) <.01 The progression of mean V d /V t over the first 28 ICU.Mean values for patients alive at 90 days are represented by the blue line and 95% confidence intervals by the blue shaded area.Mean values for patients dead at 90 days are represented by the red line and 95% confidence intervals by the red shaded area.The numbers of patients at risk in both groups are listed above the time axis.
The results of Cox regression analysis adjusted for age and SAPS II, with hazard ratios of 90-day mortality for first and second ICU-week mean P/F ratios, mean V d /V t , cut-off values for P/F ratio and V d /V t , and for BMI on admission.
The normal range of V d /V t in our cohort might be explained by the relatively high levels of PEEP used, that is, >12 cmH 2 O already T A B L E 3 detailed information on V d /V t progression provided by the large number of blood gas analyses.This enabled the characterisation of high-resolution trends V d /V t during up to 4 weeks after admission to ICU.Calculating V d /V t is feasible in the clinical setting and provides an added insight into the nature of respiratory failure in COVID-19 associated ARDS patients.The finding of a cut-off value for poor outcome below the accepted normal limit for V d /V t may appear surprising.Arguably the trend is more informative than the absolute value, and the finding of a strong association between progressively increased V d /V t during the second week of ICU admission and the 90-day mortality lends support to this view.It is also consis-