Association between Conventional Oxygen Therapy and Characteristics and Outcomes of Adult Patients in a Surgical Intensive Care Unit

Background: Oxygen therapy is often used in emergency departments and intensive care units. The prevention of hypoxia with associated complications remains the main target, but a high-concentrated usual oxygen therapy seems not to be the best strategy. We hypothesize that physiological pressure of arterial oxygen (paO 2 ) reduces mortality, onset of new infections and organ dysfunctions in critically ill patients compared with supraphysiological paO 2 . Methods: In this retrospective exploratory cohort study we included 112 critically ill adult patients treated in a surgical critical care unit. All patients were assigned to two groups dened a priori based on paO 2 mean values measured in the rst 24 hours of mechanical ventilation: rst group paO 2 75-100 mmHg (n=43), second group patients with paO 2 > 100mmHg (n=69). Primary outcome was the cumulative survival, dened from the day of admission in the intensive care unit (ICU) until death or end of the hospital stay. Secondary outcome was the incidence of infections and new organ dysfunctions in both groups. Results: The baseline characteristics like age, body mass index (BMI), lactate and severity of disease scores were similar in both groups. A total of 27 of 69 patients (39,1%) in the group paO 2 > 100mmHg and 12 of 43 patients (27,9%) in the group paO 2 75-100mmHg died during their ICU stay or further hospital stay (p=0.54). There were no statistically signicant differences in the incidence of new infections and new organ dysfunctions between the two groups. Positive end expiratory pressure (PEEP) and fraction of inspired oxygen (F i O 2 ) were in the group of patients with paO 2 > 100mmHg signicantly lower 8,4 mbar vs. 9,5 mbar (p=0.03). Conclusions: There was no signicant increase in overall mortality or new onset of infections and organ dysfunctions in critically ill adult ICU patients requiring oxygen therapy with supraphysiological paO 2 (> 100 mmHg) compared to patients with physiological paO 2 (75-100 mmHg). Further studies are needed to dene the optimal paO 2

hospital stay (p=0.54). There were no statistically signi cant differences in the incidence of new infections and new organ dysfunctions between the two groups. Positive end expiratory pressure (PEEP) and fraction of inspired oxygen (F i O 2 ) were in the group of patients with paO 2 > 100mmHg signi cantly lower 8,4 mbar vs. 9,5 mbar (p=0.03).
Conclusions: There was no signi cant increase in overall mortality or new onset of infections and organ dysfunctions in critically ill adult ICU patients requiring oxygen therapy with supraphysiological paO 2 (> 100 mmHg) compared to patients with physiological paO 2 (75-100 mmHg). Further studies are needed to de ne the optimal paO 2 parameter.

Background
Oxygen therapy is one of the primary supportive care for patients with or at risk of developing hypoxic respiratory failure admitted to the emergency department and intensive care units. Severe illness and underlying conditions can lead to low oxygen saturation in the blood, posing a risk for low tissue oxygen levels (hypoxia), ultimately leading to organ dysfunction and failure. To prevent or treat hypoxia, critically ill patients are given oxygen via oxygen masks or during mechanical ventilation in intubated patients.
Although prolonged liberal oxygen therapy was often performed to protect patients from possible organ damage caused by hypoxia, the opinion regarding oxygen therapy in the last two decades has fundamentally changed 3,5 . Numerous studies demonstrated that hyperoxia (too much oxygen) can also be harmful and is associated with unfavourable outcomes.

Page 3/14
The prevention of hypoxia and its associated complications like hypoxic brain damage further remains the main focus, but a high-concentrated liberal oxygen therapy may not be the best strategy. The minimum value for partial pressure of oxygen before organ functions are jeopardized, thus harmful for the patient, is not clearly de ned. Several studies showed that supraphysiological oxygenation with arterial pressure of oxygen (paO 2 ) of more than 75-100mmHg is associated with negative effects on the central nervous system (headache, dizziness, neuropathy), on the cardiovascular system (lower cardiac output, coronary blood ow 10 ), and may lead to hyperoxic vasoconstriction 11 on pulmonary system with atelectasis. Furthermore, the use of high inspiratory oxygen fraction causes increased oxidative stress due to radical oxygen species (ROS) 2 . The lungs are the most affected organ by oxygen toxicity. In an older study from 1985 lung damage in an intubated mouse was observed after oxygen therapy with 100% oxygen concentration after 24 hours 15 . Some studies report increased risk of mortality in critically ill patients associated with hyperoxia 5-7 .
In the presented retrospective, exploratory, monocentric study we analysed clinical outcome of intubated and mechanically ventilated adults admitted to a surgical intensive care unit. The aim of our study was to investigate supraphysiologic oxygenation, de ned as arterial partial pressure of oxygen (paO 2 ) more than 100 mmHg per 24 hours, increases mortality. Further, the incidence of organ dysfunction or failures and infections was investigated.

Study design
This retrospective exploratory cohort study was approved by the Medical Ethics Commission of the Medical Faculty of Heidelberg University, Heidelberg, Germany (S-357/2020) and conducted in accordance to ethical standards of the latest version of the Helsinki Declaration (2013) 17 .

Participants
All mechanically ventilated adult patients (≥18 years) with ≥24 hours hospital stay was screened for inclusion into the study. We excluded non-intubated patients, patients with acute respiratory distress syndrome (ARDS, according to Berlin de nition 16 ) and/or extracorporeal membrane oxygenation (ECMO), patients with severe chronic obstructive pulmonary disease (COPD GOLD IV), brain death, neutropenia and pregnant women. To avoid potential bias, re-admissions to ICU within the same hospitalization were excluded.
Overall, we enrolled 112 patients undergoing elective or emergency surgery, liver transplantation and patients with sepsis or septic shock, post-surgical complications, and haemorrhage that required an admission to the ICU. The patients in the cohort were assigned to two groups according to the paO 2 mean values: group 1 with paO 2 ≤100 mmHg, group 2 with paO 2 >100mmHg. We analysed and calculated mean values of all paO 2 values from the ABGs (arterial blood gases) measured in the rst 24 hours of mechanical ventilation for each patient. Depending on respiratory insu ciency or other causes between 9 and 21 ABG-tests were performed in 24 hours for each patient ( gure 1).

Variables
In this study we used data of patients from I-SH/ i.s.h.med (hospital electronic patient record, SAP) admitted to the surgical ICU of Heidelberg University Hospital between 1 st January and 31 st December 2018 (12 months). We searched systematically for all relevant DRGs (German Diagnosis Related Groups) that were recorded throughout the patient's stay in the ICU and that may be in uenced by hyperoxia.
Further, we compared the values of common ICU scoring systems, re ecting the severity of disease: sequential organ failure assessment (SOFA), Acute Physiology and Chronic Health Evaluation II (APACHE) and Simpli ed Acute Physiology Score -II (SAPS II) at admission and maximal value during the ICU stay. The inspired oxygen fraction (FiO 2 ), paO 2 from ABGs, positive end-expiratory pressure (PEEP), lactate averages were de ned as the mean value of all measurements within the rst 24 hours after admission. Moreover, patients' data included age, body mass index (BMI) and length of stay (LOS) in the ICU. No patient received nitrous oxide.
Primary endpoint of this study was the cumulative survival, de ned from the day of admission in the ICU till death or end of the hospital stay. Secondary outcomes were the incidence of infections and new organ dysfunctions in both groups.

Statistical methods
A comprehensive descriptive statistical analysis was conducted with mean, standard deviation, minimum, median, maximum of all continuous data and scores at admission and maximal during the stay. Binary data we reported with absolute and relative frequencies. Baseline demographics, prognostic variables, scores and ventilation modalities were compared between the groups using chi-square and Wilcoxon tests, as appropriate. Cumulative probabilities of survival were plotted using the Kaplan-Meier method and compared using log-rank test. Statistical signi cance was indicated by p-value (Pearson Chi-Square test and two-sided Wilcoxon Test) below 0.05 for baseline characteristics of patients and the secondary outcomes. SPSS-Software (IBM ® SPSS ® Statistics Version 23) and Microsoft Excel were used for the statistical data analysis.

Results
We enrolled adult patients treated on the surgical ICU of Heidelberg University Hospital from January 2018 to December 2018. 189 patients were identi ed as potential study participants. A total of 112 met our inclusion criteria and were eligible for the analysis ( gure 1). All patients were assigned to two groups: the rst group consists of patients with paO 2 75-100 mmHg (n=43), and the second group patients with paO 2 >100mmHg (n=69). The maximum value for paO 2 was 168 mmHg in two patients of group 2.

Cumulative survival
A total of 27 of 69 patients (39, 1%) in the group paO 2 >100mmHg and 12 of 43 patients (27, 9%) in the group paO2 75-100mmHg died during their ICU stay or further hospital stay. Higher paO2 was not associated with increased overall mortality. In the Kaplan-Meier survival analysis provided in gure 2 no statistically signi cant differences between groups were observed in survival estimates (p= 0.5428).

Secondary outcomes
There were no statistically signi cant between-group differences for the incidence of new infections and new organ dysfunctions (DRG Codes) occurred during the ICU stay as shown in table 1. In patients with hemorrhage (p=0.53) and polytrauma (p=0.15) as a cause for the admission to the ICU more patients received oxygen therapy resulting in paO 2 > 100 mmHg.
Both study groups had similar age, BMI, lactate, and severity of disease scores at admission and during their ICU stay, shown in baseline characteristics in table 2. Of note, PEEP and FiO 2 were in the group of patients with paO 2 >100mmHg signi cantly lower (p=0.03 and p=0.01). The patient's group with paO 2 > 100mmHg spent an average 1.1 day longer in the ICU (p=0.59).

Discussion
In this retrospective exploratory cohort study, we compared patients with physiological partial pressure of oxygen (paO 2 75-100 mmHg) and patients with supraphysiological oxygenation with paO 2 over 100mmHg. Interestingly, the highest value in the second group was 168 mmHg in two patients, which is comparably lower than paO 2 values in previous studies, in which hyperoxia was de ned as paO 2 of 300 mmHg or greater and was associated either with higher mortality 9 or with worse neurological outcome 12 .
This means that in our patient cohort, no excessive oxygenation occurred. Our results are consistent with the current systematic review with meta-analysis and trial sequential analysis that found neither bene cial nor harmful effects of higher versus lower oxygenation strategies. Similarly, the investigators found no evidence that higher oxygenation versus lower oxygenation had a profound effect on the allcause mortality 1 .
To date, there is no consensus or de nitions for cut-off values, as well as the lower and upper limit for partial pressure of oxygen, which differentiate bene cial and harmful effects for the patient. In our patient cohort with relatively small differences in paO 2 values (79-168 mmHg), it was not possible to de ne this cut-off for paO 2 . Our results differ from those in a previous single-center randomized clinical trial (RCT) by Girardis et al. 5 In that trial, lower mortality was observed in patients receiving conservative oxygen therapy compared to patients with usual oxygen therapy. The results of this trial could be in uenced by initial baseline imbalances between the study groups, such as age, severity of illness, and organ failures. Further, the study was stopped prematurely after increased mortality in one of the study groups in the interim analysis was observed, which may have resulted in an effect overestimation 4 .
Patients with haemorrhage and polytrauma as admission cause were over-represented in the paO 2 > 100 mmHg group (28% vs 35% p = 0.53 and 2% vs 10% p = 0.15, respectively). These patients often receive oxygen therapy, since clinicians tend to perform oxygen therapy more aggressively in such critical, dynamical situations with potential increased hypoxia risk aggravated by blood loss. Thus, providing a potential explanation for this discrepancy.
Our evaluation certainly presents some limitations. First, this is a single-center study and the patient cohort with 112 enrolled patients is relatively small. In order to de ne a cut-off for partial pressure of oxygen and to provide more precise and robust estimates of treatment effects, more patients needs to be enrolled to achieve an adequate power of analysis. However, the two study groups were well-balanced regarding the baseline characteristics and a detailed analysis was performed in respect to the interaction between all in the intensive care medicine relevant DRG-Codes like sepsis, septic shock, organ transplantation, delirium, prolonged weaning etc. Second, in our study, we investigated the mortality during the hospital stay without follow-up like in other studies, so that mortality in both groups may be underestimated 13,8 .
Nevertheless, our data implied that paO 2 levels of up to 168mmHg were not associated with a worse clinical outcome and that supraphysiological paO 2 levels can be tolerated without signi cant negative effects on the patients, which warrants further scrutiny.

Conclusions
In this study we did not observe any signi cant differences in mortality nor in new onset of infections and organ dysfunctions between both groups. These results suggest that paO 2 levels between 70 and 170 mmHg could be a reference point in the management of oxygen therapy to avoid excessive supraphysiological potentially toxic paO 2 values and dangerous hypoxia at the same time. Further studies with larger sample size are needed to clarify the risk-bene t ratio of pursuing a more conservative oxygen treatment strategy in critically ill patients.  Figure 1 Patients enrolment in this analysis. ARDS = acute respiratory distress syndrome; ECMO = extracorporeal membrane oxygenation; paO2 = arterial partial pressure of oxygen. NIV = non-invasive ventilation.