Alveolar-to-arterial oxygen gradient: role in the management of COVID-19 infection mild population

COVID-19 infection is frequently associated with radiological diagnosis of interstitial pneumonia and alteration in gases exchange. We decided to test arterial blood gas assay parameters, in particular alveolar-to-arterial oxygen gradient (AaDO 2 ), in predicting the need of hospitalization, the survival rate and in identifying pneumonia in patients with SARS-CoV-2 infection. We conducted an observational prospective study in one of the Emergency Department of our city. We included consecutive patients with symptoms likely related to SARS-CoV-2 infection, conrmed either with positive nasal pharyngeal swabs and/or with suggestive radiological ndings. Areas under the curve of the receiver operator characteristic curve were computed to predict need of hospitalization and the presence of pneumonia. Survival curves were analyzed using a Log-rank test. P-value less than 0.05 were considered statistically signicant. We enrolled 825 patients; the nal population was composed by 530 patients. Most of them were hospitalized due to complications, the mortality was 14% but no death occured in the ED. It results that a threshold for AaDO 2 of 27 could predict the need of hospitalization as well as a threshold for AaDO 2 of 24 could identify the presence of pneumonia. Survival curves revealed that patients with a value of AaDO 2 less than or equal to 40 had a better survival. We suggest the application of ABG parameters, in particular AaDO 2 , during the rst assessment of COVID-19 patients in the ED, because they could be additional tools to help the emergency physician to evaluate the clinical severity of patients. The study


Introduction
In December 2019 many cases of a new viral pneumonia were registered in Wuhan, China, and on 12 January 2020 WHO, the World Health Organization, isolated the genetic sequence of a new Coronavirus, named SARS-CoV-2. Then the virus has spread all over the world and on 11 March 2020 WHO declared a global pandemic 1 .
As reported in the recent literature, it could be possible to classify patients with SARS-CoV-2 infection into ve different clinical phenotypes, which implies a progressive clinical worsening from the rst to the last one 4 .
The diagnosis of the new coronavirus infection (COVID-19) is con rmed by a positive result of real time reverse transcriptase-polymerase chain reaction (RT-PCR) assay of nasal and pharyngeal swabs 5 . According to the current studies, the sensitivity of the nasopharyngeal swab is not high especially for mild cases (62,5%). For this reason, for patients with typical symptoms a negative test should be interpreted with caution and a repeat test may be considered. In case of a rst negative swab and moderate or severe symptoms, radiological typical ndings of the chest may be more sensitive than RT-PCR testing 6 .
Patients with severe infection frequently present arterial hypoxemia and develop an acute respiratory distress syndrome (ARDS) which require, in approximately 5-10% of cases, intensive care unit (ICU) admission and mechanical ventilation 7 .
While patient's oxygenation is evaluated easily by using a pulse oximeter, arterial blood gas assay (ABG) lead to a more precise measure of gas exchange: indeed with this exam it is possible to know the values of arterial partial pressure of oxygen (PaO 2 ) and arterial partial pressure of carbon dioxide (PaCO 2 ) in order to better highlight the patient's ventilatory pattern. In this way, it is possible to calculate quickly the ratio of PaO 2 to the fraction of inspired oxygen (FiO 2 ) de ned as PaO 2 /FiO 2 (P/F) and the alveolar-to-arterial oxygen gradient (AaDO 2 ) 8 the last one obtainable by the following formula: It is known that diseases that increase shunt or alter the diffusion barrier for oxygen from the alveoli to the blood can have an impact on these parameters, increasing the value of AaDO 2 and decreasing the value of P/F 9 .
As reported in literature, hypoxemia accompanied by a normal AaDO 2 and by increased PaCO 2 signi es hypoventilation, that is uncommon in COVID-19 pneumonia. Indeed, it is usually accompanied by an increased AaDO 2 , signifying either ventilation-perfusion mismatch or intra-pulmonary shunting 10 .
The primary aims of this study are to demonstrate the capability of AaDO 2 compared to P/F in predicting the need of hospitalization and the survival in patients with COVID-19 infection and phenotypes 1-4 who present to the Emergency Department 4 . The secondary aim is to evaluate usefulness of these parameters in identifying pneumonia in this cohort of patients.

Methods
We conducted an observational prospective study in the Emergency Department (ED) of Sant'Orsola University Hospital in Bologna, Italy, from 06 March 2020 to 04 April 2020. We considered 18 or older aged consecutive patients admitted to the ED with symptoms likely related to SARS-CoV-2 infection as fever, cough, dyspnea, sore throat, nasal congestion, fatigue, diarrhea, arthralgia, anosmia, ageusia. At the ED presentation, an ABG assay has been made to all these patients, independently to the real need of oxygen support.
We enrolled patients with laboratory con rmation for SARS-CoV-2 infection (positive result of nasal and pharyngeal swabs 5 ) and/or with suggestive radiological features of COVID-19 on HRTC and/or lung US 1,6 . We excluded the patients who presented to our Emergency Department with oxygen support applied by the ambulance service and so with FiO 2 > 21%.
We calculated AaDO 2 , the incremental percentage variation between the calculated and the expected aged AaDO 2 (%AaDO 2 ) and P/F for each patient using arterial blood gas analysis assessed in FiO 2 21% (ambient air) at the rst medical evaluation in the ED. According to the literature, we considered patients with AaDO 2 > 5 mmHg 10,11,12 . Patients were followed up until 11 May 2020 recording ED discharge, the need of hospitalization and/or death through data from the hospital records.
Demographic, clinical and laboratory characteristics of the cohort are expressed as means and standard deviation (SD) and as number and percentage as appropriated. Areas under the curve (AUCs) and the 95% con dence interval (CI) of the receiver operator characteristic (ROC) curve were computed to predict need of hospitalization, presence of viral pneumonia and mortality. The optimal AaDO 2, %AaDO 2 and P/F index cut-offs were determined optimizing sensitivity and speci city, privileging sensitivity for our purposes. Survival curves were analyzed using a Log-rank test.
P-value less than 0.05 were considered statistically signi cant. The analysis was obtained using IBM SPSS Statistics software version 25.
The study was approved by our local ethics committee with the number 551/2020/Oss/AOUBo.

Results
We enrolled 825 consecutive patients admitted to the ED of Sant'Orsola University Hospital in Bologna, Italy. Of these patients, 107 were excluded because of negative swab and radiological ndings not suggestive for SARS-CoV-2 pneumonia, 58 were excluded because of a different discharge diagnosis, 61 because they arrived in ED with oxygen support (FiO 2 > 21%) and 69 because of AaDO 2 < 5 mmHg. So, the nal study group was composed by 530 patients ( Figure 1). Table 1 shows demographic characteristics and comorbidities. Overall, 55.1% (292 of 530 patients) were male, the mean age was 62.5 years old and the median age was 61 years old. Main comorbidities were: hypertension, diabetes, chronic kidney disease (CKD), chronic coronary artery disease (CAD), chronic obstructive pulmonary disease (COPD), asthma and active malignancy. Table 2 shows ABG parameters at the admission in the ED. The pH range varies from 7.20 to 7.64 (mean 7.46); the paO 2 varies from 26 mmHg to 118 mmHg (mean 72 mmHg); the paCO 2 varies from 15 mmHg to 73 mmHg (mean 33 mmHg). The mean value of AaDO 2 was 36 (range 5-83); the mean value of %AaDO 2 was 85% (range 78-384); the mean value of P/F was 344 (range 124-562). The predictive values of AaDO 2 , %AaDO 2 and P/F obtained by means of ROC curves are listed in Table 3.
Out of 168 patients with AaDO 2 less than or equal to 27, only 3 (1.8%) were readmitted in the ED within 7 days and subsequently hospitalized. Of these 3 patients, 2 did not have pneumonia on rst chest imaging and, during the second assessment, they showed a worsening of AaDO 2 and P/F values above 27 and under 365 respectively and development interstitial pneumonia. curve showed that the AaDO 2 threshold value of 40 (70% sens. and 64% spec.) and the P/F threshold value of 300 (78% sens. and 62% spec.) corresponded to acceptable cut off for predicting mortality.
Survival curves are shown in Figure 3. Log-rank test revealed statistically signi cant differences (p < 0.0001) between the population with AaDO 2 less than or equal 40 (23 deaths of 312 patients) and with AaDO 2 more than 40 (55 deaths of 218 patients). Log-rank test revealed also statistically signi cant differences (p < 0.0001) between the P/F less than or equal to 300 population (48 deaths of 151 patients, 31.8%) and P/F more tham 300 population (30 deaths of 379 patients, 7.9%).

Discussion
Considering the impact of COVID-19 interstitial pulmonary involvement on gases exchange, ventilatory/perfusion mismatch and shunting, we decided to test ABG parameters, in particular AaDO 2 , %AaDO 2 and P/F, to evaluate the e cacy of this tool to better guide ED physicians in this new challenging clinical condition. Arterial blood gas assay is easily available in an emergency setting and it gives immediately crucial information about pulmonary involvement and respiratory dynamics. Moreover, we decided to use the alveolar-to-arterial oxygen gradient instead of PaO 2 because of its better capability to underline the presence of ventilatory/perfusion mismatch and shunting 9 .
In our cohort, most patients were male and the median age was 61 years old, similarly to that reported for other Italian COVID-19 cases based on data sourced from Epicentro Epidemiology for public health (Istituto Superiore di Sanità https://www.epicentro.iss.it/en/coronavirus/sars-cov-2-dashboard).
Of 530 enrolled patients, about 72% were hospitalized because of pneumonia diagnosis and based on clinical gestalt. In our study group, the mortality rate (14%) was similar to that recently reported in Italy 13 and no death was recorded at the time of discharge to the emergency room. Only 23% of the patients required oxygen therapy and no one was intubated during the rst evaluation in the ED because they showed initial good responsiveness to the oxygen support, implying the presence of ventilatory-perfusion mismatch instead of intra-pulmonary shunting 10 . Moreover, these data could be justi ed by the characteristics of enrolled patients, who, in most cases, presented with mild symptoms and without need of respiratory support before ED evaluation and a subgroup of them could be otherwise managed in an outpatient setting and not necessarily in the ED.
From statistical analysis, it emerged that in our mild population the AaDO 2 threshold value of 27 could have an higher sensitivity than P/F threshold of 365 in predicting hospital admission after ED evaluation (respectively 87% vs 80%). As far as the prediction of pneumonia is concerned, the AaDO 2 threshold of 24 is better related to the diagnosis of pneumonia in our COVID-19 cohort's patients than P/F. Furthermore, as already known in literature in other clinical conditions, AaDO 2 more than 20 is associated with most serious clinical features that require hospitalization 7 .
In order to enhance the effect of AaDO 2 regardless of the age, we tested the incremental percentage variation between the calculated and the aged expected AaDO 2 , but, despite of its statistical signi cance, it showed poor performances, probably due to the impact of the age itself on patient's COVID prognosis. As a consequence, in clinical practice, we suggest better to use the absolute value of AaDO 2 instead of its incremental percentage.
From the analysis of the survival curves, it results that patients with value of AaDO 2 less than or equal 40 and P/F more than 300 had a better survival, according to the literature in which severe lung injury occurs with highest value of AaDO 2 and worst P/F value 14 .
The main limitation of this study is the relatively small cohort, composed by patients with "mild" infection, who are classi able in the rst four clinical phenotypes 4 . We commit to validate and eventually con rm the data obtained in this study enlarging the study group.
Of our best knowledge, this is one of the few studies published until now which analyses patients with SARS-CoV-2 infection in the ED and the rst one in which ABG parameters are discussed with particular focus on the role of AaDO 2 .
We suggest the application of ABG parameters, in particular AaDO 2 , during the rst assessment of COVID-19 patients in the ED, because they could be additional tools to help the emergency physician to evaluate the clinical severity of patients.

Declarations
Funding: none Con icts of interest/Competing interests: none Ethics approval: study approved by our local ethics committee with the number 551/2020/Oss/AOUBo Consent to participate: informed consent was obtained from all individual participants included in the study.

Consent for publication: not applicable
Availability of data and material: the datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.