Oxygen metabolism markers as predictors of mortality in severe COVID-19

Abstract

Objective

To investigate the use of oxygen metabolism markers as predictors of mortality in patients with severe coronavirus disease 2019 (COVID-19).

Methods

A retrospective analysis was undertaken to compare the medical records of patients with severe COVID-19 (53 deceased patients and 50 survivors). The survivors were selected from 222 records using a random number generator. In addition, 28 individuals who considered themselves to be healthy and who had no history of serious illness were included in the study for comparison. Oxygen saturation in arterial blood, oxygen saturation in central venous blood (ScvO₂), arterial partial pressure of oxygen (PaO₂), respiratory index (PaO₂/fraction of inspired oxygen), oxygen delivery, oxygen consumption (VO₂) and oxygen extraction (O₂ER) were compared in all participants. The optimal cut-off point for each oxygen metabolism marker in the prediction of mortality was determined based on the maximum value of the Youden Index in receiver operating characteristic curve analysis.

Results

Significant differences in all studied oxygen metabolism markers were found between survivors compared with deceased patients (p < 0.001). ScvO₂, VO₂ and O₂ER [area under curve (AUC) 1.0] were the strongest predictors of mortality, and PaO₂ was the weakest predictor of mortality (AUC 0.81). ScvO₂ <29%, VO₂ >124.6 ml/min and O₂ER >30.2% were identified as predictors of mortality in patients with COVID-19.

Conclusion

ScvO₂, VO2 and O₂ER are good predictors of mortality in critically ill patients with COVID-19.

Introduction

Recently, the medical world has focused its attention on the diagnosis and treatment of coronavirus disease 2019 (COVID-19) (Coronavirus (COVID-19), 2020). The number of cases of COVID-10 and the number of associated deaths are increasing daily. Due to the significant number of deaths, prediction of the outcome of COVID-19 is essential, and there is a need to identify predictive markers of mortality for infected patients. This was the subject of a recent meta-analysis by Tian et al. (2020), who reported that levels of cardiac troponin, C-reactive protein (CRP), interleukin-6, D-dimer, creatinine, alanine transferase and albumin can be used to predict mortality in patients with COVID-19. In another meta-analysis, Henry et al. reported that the absolute values of lymphocytes, platelets, albumin, total bilirubin, urea, creatinine, myoglobin, cardiac troponin, CRP and interleukin-6 were potential predictors of mortality in patients with COVID-19 (Henry et al., 2020). As far as is known, no studies to date have investigated whether oxygen metabolism markers can be used to predict mortality in patients with COVID-19. The most common indices used to estimate the severity of respiratory failure in patients with COVID-19 are arterial oxygen saturation (SaO₂), partial pressure of oxygen in arterial blood (PaO₂) and the respiratory index [PaO₂/fraction of inspired oxygen (FiO₂)] (Coronavirus (COVID-19), 2020). Considering the high mortality rate in cases of severe COVID-19, there is an urgent need to identify those patients at increased risk of death. Early intensification of treatment in this group is crucial. This study investigated the use of oxygen metabolism markers as predictors of mortality in patients with severe COVID-19.

Methods

Study design

This retrospective observational study analysed the medical records of patients with severe COVID-19 [i.e. interstitial pneumonia with acute respiratory distress syndrome (ARDS) and acute respiratory insufficiency] treated in Kiev City Clinical Hospital No. 4 between 2 February 2020 and 15 September 2020. ARDS was defined using the Berlin definition (Costa and Amato, 2013).

Selection of participants

The inclusion criteria used to select patients with COVID-19 were:

• •severe acute respiratory syndrome coronavirus-2 infection confirmed by reverse transcription polymerase chain reaction;
• •presence of diffuse, bilateral lung inflammation on computed tomography; and
• •PaO2/FIO2 ratio <200.

The exclusion criteria for patients with COVID-19 were:

• •the presence of comorbidities that could have caused death (i.e. cardiogenic pulmonary oedema, advanced chronic pulmonary disease, active malignancy, pulmonary embolism, diabetic ketoacidosis, advanced chronic kidney diseases, pregnancy, brain stroke and myocardial infarction); and
• •participation in other clinical studies.

In total, 272 medical records that met the study criteria were identified through initial screening. Among these individuals, 53 patients (28 female, 52.8%) had died. The remaining 222 medical records were numbered using a random number generator (Costa and Amato, 2013), and 50 (23 female, 46%) patients were selected to represent survivors. In addition, 28 (10 female, 35.7%) individuals who considered themselves to be healthy, who had no history of serious illness, and who were awaiting ophthalmic surgery were included in the study for comparison.

An overview of basic data for the study population is presented in Table 1 .

Table 1.

Baseline parameters and laboratory test results of study patients.

	Healthy subjects, n = 28	COVID-19 survivors, n = 50	COVID-19 deceased patients, n = 53	p-value for correlation between examined groups
				Healthy vs survivors	Healthy vs deceased	Survivors vs deceased
Age, years, mean (SD), range	66.3 (4.3), 55–77	70.5 (4.2),61–80	67.8 (4.0), 59–78	<0.001	0.29	<0.001
Temperature, mean (SD), °C	36.5 (0.1)	38.0 (0.2)	38.0 (0.2)	<0.001	<0.001	1.0
Systolic blood pressure, mean (SD), mmHg	133 (14)	132 (15)	133 (14)	0.93
Diastolic blood pressure, mean (SD), mmHg	83 (9)	82 (9)	83 (9)	0.90
Creatinine, mean (SD), mmol/l	0.09 (0.02)	0.11 (0.02)	0.13 (0.11)	<0.001	<0.001	0.99
C-reactive protein, mean (SD), mg/l	3.80 (0.63)	47.64 (12.83)	43.94 (13.78)	<0.001	<0.001	0.68
Procalcitonin, mean (SD) (ng/ml)	0.19 (0.03)	1.28 (0.45)	1.19 (0.39)	<0.001	<0.001	1.0

COVID-19, coronavirus disease 2019; SD, standard deviation.

Measurement methods

Arterial blood was sampled from the radial artery and venous blood was sampled from the internal jugular vein during catheterization. In patients with COVID-19, sampling was performed immediately after admission to the intensive care unit. SaO₂, oxygen saturation in central venous blood (ScvO₂), PaO₂, PaO₂/FiO₂, oxygen delivery (DO₂), oxygen consumption (VO₂) and oxygen extraction (O₂ER) were compared in all participating individuals.

A BGA 101 gas analyser (Wondfo, Guangzhou, China) was used to measure PaO₂, SaO₂ and ScvO₂. The cardiac index was estimated using a portable non-invasive cardiometer (ICON; Cardiotronic, Inc., La Jolla, CA, USA).

DO₂ (ml/min) was calculated as:

DO₂ = 1.34 × SaO₂ × CO × Hb / 100where 1.34 is Huffner's constant, Hb is the

blood haemoglobin concentration (g/l), SaO₂ is arterial oxygen saturation (%), CO is cardiac output (l/min), and 100 is the unit conversion index.

VO₂ (ml/min) was calculated as the difference between arterial and venous oxygen transport (Marino, 2013):

VO₂ = CO × Hb × 1.34 × (SaO₂ – ScvO₂) /100where CO is

cardiac output (l/min); Hb is haemoglobin concentration (g/l); 1.34 is Huffner’s constant; SaO₂ and ScvO₂ are oxygen saturation in arterial blood and oxygen saturation in central venous blood, respectively (%); and 100 is the unit conversion index.

FiO₂ was calculated as:

FiO₂% = 20 + (4 × O₂ l/min)w

here O₂ is the oxygen supply speed.

O₂ER was calculated as (Marino, 2013):

O₂ER = VO₂ / DO₂ x 100%.

Statistical analysis

Statistical analysis was undertaken using Statistica Version 13.1 (TIBCO Software Inc., Palo Alto, CA, USA). Non-parametric statistics were used to compare categorical variables between the study groups. Demographics and laboratory results for the three groups (healthy patients, survivors and patients who died due to COVID-19) were compared using the post-hoc Kruskal–Wallis test. The optimal cut-off point for each oxygen metabolism marker for predicting mortality was determined based on the maximum value of the Youden Index in receiver operating characteristic curve (ROC) curve analysis. For all statistical tests, p < 0.05 was considered to indicate significance.

Results

This study found that patients with COVID-19 had a significantly higher temperature, and CRP, procalcitonin and creatinine levels (p < 0.001) compared with healthy subjects (Table 1).

Oxygen metabolism indices in patients with COVID-19 differed significantly from those in healthy subjects, and also between patients with COVID-19 who survived and those who died (Table 2 ). All patients with COVID-19 had significant oxygen metabolism disorders which were manifested by substantial decreases in SaO₂, PaO₂, PaO₂/FiO₂ and DO₂. SaO₂ in survivors and deceased patients was 2.16 and 2.42 times lower, respectively, compared with healthy subjects, and 1.12 times lower in deceased patients compared with survivors (4.88% lower). Similarly, PaO₂ in survivors and deceased patients was 2.97 and 3.22 times lower, respectively, compared with healthy subjects, and 1.08 times lower in deceased patients compared with survivors (2.5 mm lower). DO₂ in survivors and deceased patients was 2.15 and 2.41 times lower, respectively, compared with healthy subjects, and 1.12 times lower in deceased patients compared with survivors (46.44 ml/min lower). PaO₂/FiO₂ in survivors and deceased patients was 3.13 and 3.39 times lower, respectively, compared with healthy subjects, and 1.08 times lower in deceased patients compared with survivors (11.59 mmHg lower).

Table 2.

Values of oxygen metabolism markers in study patients.

	Healthy individuals, n = 28	COVID-19 survivors, n = 50	COVID-19 deceased patients, n = 53	p-value for correlation between examined groups
				Healthy vs survivors	Healthy vs deceased	Survivors vs deceased
SaO2, mean (SD), %	97.07 (0.98)	44.90 (2.06)	40.02 (3.03)	<0.001	<0.001	<0.001
ScvO2, mean (SD), %	66.07 (3.05)	33.18 (1.93)	17.94 (1.64)	<0.001	<0.001	<0.001
PaO2, mean (SD), mm Hg	95.36 (3.15)	32.14 (1.70)	29.64 (1.99)	<0.001	<0.001	<0.001
PaO2/FiO2, mean (SD), mm Hg	475.71 (16.03)	152.12 (3.73)	140.53 (5.49)	<0.001	<0.001	<0.001
DO2, mean (SD), ml/min	905.90 (39.39)	421.99 (18.95)	375.55 (23.87)	<0.001	<0.001	<0.001
VO2, mean (SD), ml/min	281.75 (11.29)	112.18 (4.95)	206.02 (15.31)	<0.001	<0.001	<0.001
O2ER, mean (SD), %	31.16 (1.88)	26.51 (1.49)	54.89 (1.53)	<0.001	<0.001	<0.001

COVID-19, coronavirus disease 2019; SaO₂, oxygen saturation in arterial blood; SD, standard deviation; ScvO₂, oxygen saturation in central venous blood; PaO_2, arterial partial pressure of oxygen; FiO₂, fraction of inspired oxygen; PaO₂/FiO_2, respiratory index; DO₂, oxygen delivery; VO₂, oxygen consumption; O₂ER, oxygen extraction.

Analysis was also conducted for ScvO₂, VO₂ and O₂ER. ScvO₂ in survivors and deceased patients was 1.99 and 3.68 times lower, respectively, compared with healthy subjects. ScvO₂ in survivors was 1.85 times higher compared with deceased patients (15.24 mmHg higher). VO₂ in deceased patients was 1.84 times higher compared with survivors (93.84 ml/min higher). Similarly, O₂ER in deceased patients was 1.76 times higher compared with healthy subjects, and 2.07 times higher compared with survivors. O₂ER in deceased patients was 1.76 times higher compared with survivors (28.38% higher).

All of the differences reported above were statistically significant (p < 0.001).

A discrimination model was established to determine the values of oxygen metabolism markers for predicting mortality. ROC analysis was used to calculate the cut-off points (Table 3 ). The analysis revealed that all parameters had prognostic value, with ScvO₂, VO₂ and O₂ER [area under curve (AUC) 1.0] being the strongest predictors of mortality, and PaO₂ being the weakest predictor of mortality (AUC 0.81).

Table 3.

Performance of oxygen metabolism markers for predicting death using logistic regression analysis.

	Cut-off point	AUC	95% CI	p-value	Youden Index
SaO2, %	43	0.94	0.90–0.98	<0.001	0.75
ScvO2, %	29	1	1	<0.001	1.0
PaO2, mmHg	31.6	0.81	0.73–0.89	<0.001	0.43
PaO2/FiO2, mmHg	144.5	0.96	0.93–0.99	<0.001	0.79
DO2, ml/min	401	0.95	0.92–0.99	<0.001	0.83
VO2, ml/min	124.6	1	1	<0.001	1.0
O2ER, %	30.2	1	1	<0.001	1.0

AUC, area under curve; CI, confidence interval; SaO₂, oxygen saturation in arterial blood; ScvO₂, oxygen saturation in central venous blood; PaO_2, arterial partial pressure of oxygen; FiO₂, fraction of inspired oxygen; PaO₂/FiO_2, respiratory index; DO₂, oxygen delivery; VO₂, oxygen consumption; O₂ER, oxygen extraction.

Discussion

A limited number of studies on hypoxia have been undertaken in patients with COVID-19. SaO₂, PaO₂ and PaO₂/FiO₂ are most often used to characterize the degree of respiratory insufficiency in patients with COVID-19 (Coronavirus (COVID-19), 2020). Li et al. (2020) studied the pathogenesis of COVID-19, and stated that a severe form of the disease progresses into sepsis and ARDS, and consequently into severe hypoxia. The latter is the leading cause of death in these patients. Xie et al. (2020) suggested that hypoxaemia in COVID-19 is predictive of mortality. In their opinion, careful monitoring of oxygenation helps in the clinical management of patients with severe COVID-19, especially if limited intensive care resources are available Xie et al. (2020).

ScvO₂ measurements provide insight into the balance between oxygen supply and tissue oxygen demand. Physiologically, ScvO₂ is in the range of 65–75% and usually exceeds 70% (van Beest et al., 2011). A decrease below 70% is evidence of tissue hypoperfusion (Walley, 2011). A decrease in ScvO₂ can be caused by tissue hypoperfusion, arterial desaturation or a decline in haemoglobin concentration. In critical conditions, dynamic changes in ScvO₂ are more significant than changes in SaO₂ (Jones et al., 2010, Smetkin and Kirov, 2018).

ScvO₂ values can differ considerably in various clinical situations. Patients with chronic heart failure may have ScvO₂ as low as 65% without signs of tissue hypoxia due to a compensatory increase in O₂ER in response to reduced DO₂ (Nebout and Pirracchio, 2012). In patients with respiratory insufficiency, ScvO₂ is one of the oxygen metabolism markers used to set the parameters for mechanical ventilation and other respiratory treatment (Peyrony et al., 2019). A study conducted in a multidisciplinary intensive care unit showed that mortality in patients with ScvO₂ <60% was 1.7 times higher compared with patients with higher ScvO₂ values. Treatment attempts only resulted in a slight increase in ScvO₂, and did not affect the fatal outcome (Salem et al., 2019). Similar clinical findings were observed in the present study in deceased patients. Mean ScvO₂ values in deceased patients were two times lower compared with survivors, and more than three and half times lower compared with healthy subjects (Table 2). Therefore, ScvO₂ <29% appears to be predictive of mortality in patients with severe COVID-19 (Table 3). This parameter is particularly useful as it can be measured quickly and easily in all patients.

DO₂ is another marker of life support mechanism, and DO₂ disorders are crucial factors determining mortality in intensive care units (Pappachan et al., 2019). This is consistent with the present findings. In this study, DO₂ was substantially lower in patients with COVID-19 compared with healthy subjects, and the values in deceased patients were significantly lower compared with survivors (Table 2). A considerable decrease in DO₂ in both COVID-19 groups should be referred to as ARDS, the leading pathology in the study population (Pappachan et al., 2019). Pathology in DO₂ is particularly important in critically ill patients (i.e. when oxygen metabolism in the tissues is disturbed). Under normal conditions, VO₂ does not depend on DO₂. In healthy adults at rest, the body uses approximately 25% of the delivered O₂ (Walton and Hansen, 2018) (i.e. approximately 220–250 ml O₂/min). In critical conditions, VO₂ is considerably greater. An increase in body temperature by just 1 °C increases VO₂ by 10%. VO₂ is 1.5–2.0 times higher in patients with chills, and 2.0–2.5 times higher in patients with sepsis (Muzdubayeva, 2016).

DO₂/VO₂ balance is achieved by metabolic autoregulation of cells, resulting in enhanced O₂ER when DO₂ is markedly reduced (Nevares et al., 2017). This mechanism has its limits and can fail in critical conditions (i.e. when critically reduced DO₂ influences VO₂). This was observed in the COVID-19 patients in the present study, as the decrease in DO₂ also reduced VO₂. However, VO₂ in deceased patients was nearly twice as high compared with survivors (Table 2). This was likely related to an oxygen debt resulting from critical tissue hypoxia (Navalta et al., 2018). This is known as the ‘oxygen paradox’, where energy exchange disorders begin before DO₂ is reduced to a critical level (i.e. when VO₂ is proportional to supply), and can happen before the occurrence of an oxygen debt (Moen and Stuhr, 2012).

Hypoxia in patients with severe COVID-19 is determined not only by the DO₂/VO₂ ratio but also by hypoxaemic processes at subcellular, cellular, tissue and organ levels (Bhatraju et al., 2020). It is difficult to explain the higher VO₂ values in deceased patients compared with survivors. Physiologically, VO₂ depends on tissue needs alone, not on DO₂, as DO₂ exceeds tissue demands. In certain clinical circumstances, VO₂ increases in direct proportion to DO₂ (Place et al., 2017). This is known as ‘pathological dependence of VO₂ on DO₂’. Clinical observations have confirmed this pathology in patients with sepsis, where microcirculation disorders occur and VO₂ may increase; this is an extremely unfavourable sign (Dietz et al., 2019, Kirov, 2014). The present findings were similar. In survivors, the decrease in DO₂ was followed by a proportional decrease in VO₂. This was not observed in deceased patients, in whom a substantial decrease in DO₂ was accompanied by a relatively small decrease in VO₂. The abnormalities observed in O₂ER mirror tissue hypoxia. This results in multiple organ dysfunction. Xie et al. reported that therapeutic attempts to reduce VO₂ are key factors in the successful treatment of patients with COVID-19 (Xie et al., 2020). Increased VO₂ was the cause of increased hypoxia in deceased patients in the present study.

Evaluation of the imbalance between DO₂ and VO₂ can be crucial for tailoring therapy in patients with severe COVID-19, as it enables early identification and assessment of the severity of global body dysoxia. The body launches several compensatory mechanisms in response to the imbalance between DO₂ and VO₂, including increased cardiac output, increased O₂ER, and redistribution of blood flow to organs and tissues with the highest oxygen demands (Chu et al., 2018). VO₂ depends on oxidative phosphorylation activity and functional activity of the tissue at a given time. This process is characterized by O₂ER (Li et al., 2020). At rest, O₂ER is 20–30%. It is believed that one of the reasons for an increase in O₂ER from blood is the disturbance of microcirculation in the tissues (Li et al., 2019). In the present study, O₂ER values in healthy subjects and survivors, although significantly different, were close to normal ranges, but were almost twice as high in deceased patients compared with healthy subjects (Table 2). In the authors’ opinion, O₂ER >30% can be considered a good predictor of mortality in patients with COVID-19 (Table 3). The increase in O₂ER likely results from increased VO₂, but it was not possible to confirm this in the present study. Explanation of this pathology may be of critical importance for understanding the cellular pathomechanisms in severe COVID-19. Further clinical trials are needed to clarify this phenomenon.

Conclusions

Monitoring oxygen metabolism allows identification of patients with severe COVID-19. ScvO₂ <29%, VO₂ <125 ml/min and O₂ER >30% appear to be good predictors of mortality. In patients with severe COVID-19, markers of internal respiration seem to be better predictors of mortality than markers of external respiration. Further clinical studies are needed for better elucidation of these findings.

Conflict of interest

None declared.

Funding source

This study is part of the clinical research of the Department of Anaesthesiology and Intensive Care of Kyiv National Medical University on the topic ‘Optimization of respiratory support methods for patients with severe forms of respiratory insufficiency, including acute respiratory distress syndrome’ registered in the Clinical Trials Register of the State Expert Centre of the Ministry of Healthcare of Ukraine (Registration No. 0112U001413). This work was undertaken according to the subject register in Simple System SUB.A120.19.036, and was supported by statutory subvention granted by the Ministry of Science and Higher Education in Poland.

Ethical approval

This study was conducted in accordance with the ethical principles of the Declaration of Helsinki. The study was approved by the Bioethics Committee of Kiev City Clinical Hospital No. 4 (Decision No. 64, 2 July 2020). Informed consent was obtained for the control group of healthy subjects. As the study of the patients with COVID-19 was of a retrospective nature, their informed consent was not required under Ukrainian law.