Evidence for Oxygen Use in the Hospitalized Patient: Is More Really the Enemy of Good?

Early Oxygen Use

Following Priestley's published findings, Antoine Lavoisier repeated his and Scheele's experiments and proved that oxygen was a chemical element, disproving the phlogiston theory.² In 1778 he named the gas oxygen “acid former,” due to his belief that it was a component of all acids. Five years later a French physician treated a patient suffering from tuberculosis with daily inhalations of oxygen, which is believed to be the first medical use of the gas.¹ Throughout most of the 19th century, pure oxygen therapy was not available to the public. Mostly diluted nitric oxide, “compound oxygen,” was widely believed to be a panacea for many common ailments. George Holtzapple is credited with publishing the first case report describing the administration of intermittent oxygen therapy, to a 16-year-old male with lobar pneumonia, at York Hospital in 1885. The patient's cyanosis improved with oxygen therapy and he subsequently recovered.³ It was not until 1890 that Albert Blodgett administered continuous flow oxygen to a patient with pneumonia to relieve shortness of breath.⁴ He estimated that around 200 gallons of oxygen per day was needed for continuous administration: approximately 6 L/min.

Modern Oxygen Therapy

The understanding of therapy and physiology advanced quickly during the early 1900s, due to the gas poisonings during World War I, and advances in basic science.¹ Physiologists Adolph Fick and Paul Bert further advanced oxygen physiology by describing oxygen in units of partial pressure, which led to the understanding of the differences between arterial and venous blood oxygenation and the relationship to cardiac output and oxygen consumption. John Haldane published the first paper on the rational use of oxygen in 1917.⁵ Much of what we consider to be the basic physiologic concepts of oxygenation can be attributed to Haldane. In his paper he describes the respiratory drive as regulated by carbon dioxide, the different types or causes of hypoxemia, and tissue hypoxemia in carbon monoxide (CO) poisoning. He further describes the mechanisms of ventilation-perfusion matching and mismatching and the role of supplemental oxygen as a treatment. Haldane was also the first to describe the effects of oxygen on the pulmonary system.

Oxygen use on the battlefield was first reported during World War I, primarily for the treatment of phosgene gas poisoning.⁶ When mixed with water in the lungs, phosgene forms hydrochloric acid, damaging alveolar lining, and at high doses leads to pulmonary edema and eventually to what we know today as ARDS. Oxygen was also used in the treatment of trench nephritis, acute bronchitis, and severe hemorrhage. Oxygen treatment on the battlefield was accomplished by the use of equipment developed by Haldane, which consisted of a pressurized cylinder, pressure regulator, a reservoir, and mask, much like what is currently used today. Experiences learned from the war helped develop a basic understanding of rational oxygen use, ways to administer, and what did not work: mainly intermittent usage in a hypoxic patient. Evidence for oxygen use in trauma care was also gained from the war experience. Despite evidence of the benefit of continuous therapy on the battlefield and the publishing of Haldane's book, Respiration,⁷ many physicians continued to prescribe intermittent oxygen therapy into the first half of the 20th century.

Indications for Oxygen Therapy

Supplemental oxygen is an important part of modern medical care. From prehospital to in-hospital care and anesthesia applications, to long-term usage in chronic lung disease, oxygen use has become so common that it is often taken for granted. Although it is considered a drug and should be prescribed as such, oxygen is often given to patients at the caregiver's whim, and frequently without a physician's order.^8,9 This occurrence in the hospital setting is common because oxygen is readily available, abundant, and cheap when employing the large liquid systems, as do most hospitals. Even after a century's experience and numerous publications concerning oxygen administration, the question remains: what are the evidence-based indications for oxygen therapy in hospitalized patients?

The American Association for Respiratory Care provides guidance for in-hospital use of oxygen other than with mechanical ventilators and hyperbaric chambers.¹⁰ The recommended indications are documented hypoxemia (P_aO2 < 60 mm Hg or S_aO2 < 90%), suspected hypoxemia, severe trauma, acute myocardial infarction, and short-term therapy such as post-anesthesia recovery or surgical intervention. The British Thoracic Society's^11,12 and Western Australian Hospital's indications for supplemental oxygen are to maintain normal or near normal S_pO2 (94–98%) for all patients not at risk for hypercapnic respiratory failure, and S_pO2 of 88–92% for those at risk. This guidance specifically states that patients suffering from myocardial infarction and acute coronary syndrome have the same S_pO2 targets as above. Additionally, the guidance states that non-hypoxic breathless patients (other than CO poisoning) do not benefit from oxygen therapy and does not recommend supplementation. Both of the latter associations recommend the use of an oxygen alert card for those patients at risk for hypercapnic respiratory failure, so that in an emergency the appropriate low F_IO2 will be administered.

The remainder of this paper will detail the diseases/conditions for which oxygen is often prescribed as a treatment, and the available evidence to support or refute its use.

Oxygen Myths

John Downs presented the 2002 Donald F Egan Scientific Lecture entitled “Has Oxygen Administration Delayed Appropriate Respiratory Care? Fallacies Regarding Oxygen Therapy.”¹³ Downs outlined what he believed to be 3 commonly held beliefs and the related evidence regarding oxygen therapy.

Chronic Obstructive Pulmonary Disease

The World Health Organization estimates that there are 210 million people worldwide living with COPD, making it a major health issue in many countries.²¹ It is estimated that in the United States alone the cost to manage patients with the disease exceeded $49 billion in 2010.²² An additional $73 billion was associated with hospital admissions. Although oxygen is among the standard management treatments, it was shown over 50 years ago that high F_IO2 increases blood carbon dioxide concentration in some COPD patients.²³ The British Thoracic Society recommends that until an arterial blood gas is obtained, any patient with known or suspected COPD not be given an F_IO2 > 0.28.²⁴

Denniston et al conducted a prospective audit of 97 subjects with the diagnosis of COPD admitted to the emergency department, representing 101 episodes of COPD exacerbation.²⁵ At some point in the pre-hospital or emergency department setting, 56% received an F_IO2 > 0.28. For those subjects who received an F_IO2 > 0.28, in-hospital mortality was 14% (8 of 57) versus 2% (1 of 44) for those who received an F_IO2 ≤ 0.28. Demographics and smoking history were not different between the 2 groups. Interestingly, in the ambulance those subjects who either self-identified or were identified by the crew as having COPD received a mean F_IO2 of 0.47, versus 0.6 if they were not identified. Although the ambulance crew administered a lower F_IO2 to those subjects identified as having the diagnosis of COPD, it was still well above the recommended F_IO2.

In a prospective study including 972 subjects admitted to the emergency department with the diagnosis of COPD, Plant et al found that 20% had respiratory acidosis.²⁶ In 47% of the hypercapnic subjects, pH was inversely related to P_aO2, with most being associated with a P_aO2 > 75 mm Hg. As in the aforementioned study by Denniston, the acidotic subjects had a higher in-hospital morality than the non-acidotic subjects (12.8% vs 6.9%).

A recent study comparing high flow with titrated oxygen administration in the pre-hospital setting in 405 subjects with COPD (214 confirmed) was conducted by Austin and colleagues.²⁷ Subjects were randomized into 2 groups: oxygen via nasal cannula titrated to S_pO2 of 88–92%, or 8–10 L/min of oxygen via non-rebreathing mask. Both groups received standard of care bronchodilator treatments enroute to the hospital. The study results showed that titrating oxygen to maintain an S_pO2 of 88–92% reduced the risk of death from hypercapnia and respiratory failure by 58% in all subjects, and by 78% in those with confirmed COPD. In the high flow oxygen group the number needed to harm was 14.

Infants/Neonates

François Chaussier used oxygen in attempts to revive what he termed “near dead” infants, beginning in 1780,²⁸ but it was not until the 1930s that physicians began using oxygen routinely with neonates.²⁹ In 1938, Chapple reported delivering F_IO2 of approximately 0.46 to an incubator to treat preterm infants.³⁰ A decade later, Terry documented over 100 cases of a new type of blindness present in premature infants,³¹ but it was not until 1951 that Campbell³² linked supplemental oxygen to the cause of what was initially termed retrolental fibroplasia and is currently known as retinopathy of prematurity (ROP). The first randomized controlled trial (RCT) to study the association of supplemental oxygen with ROP was published in 1952 by Patz et al.³³ The infants were randomized to either an F_IO2 of 1.0 or titrated oxygen to treat hypoxemia. In the F_IO2 1.0 group, 61% of the infants developed ROP, versus 16% in the titrated oxygen group.

In a retrospective study of risk factors for developing ROP in 2009, Hua et al³⁴ found that infants that breathed F_IO2 of > 0.8 for any length of time, and those who received any supplemental oxygen for > 8 days had the highest incidence. Additionally, the lower the birth weight, the higher the incidence of ROP for those infants who received oxygen. This study concurred with a 1977 study by Kinsey et al, finding that birth weight < 1,200 g and length of exposure to oxygen increased ROP risk.³⁵

Early studies that confirmed the association between oxygen and ROP helped to increase awareness of the problem, but, due to lack of ability to continuously measure arterial oxygenation, many premature infants were left profoundly hypoxic, for fear that administering oxygen would lead to ROP. The resulting hypoxia led to increased incidence of cerebral palsy. An early paper from 1961³⁶ reported that, in a study of 1,080 premature infants, supplemental oxygen administration for < 2 days showed a 17% increase in cerebral palsy, whereas oxygen exposure for > 10 days resulted in a 22% increase in ROP. This was the first study to show that there can be neither too little nor too much oxygen given to premature infants. In a multicenter RCT involving 358 preterm infants, Askie et al showed no difference in growth and development in those infants with S_pO2 of 91–94% than those with S_pO2 of 95% and above.³⁷

The use of oxygen in the resuscitation of infants in the delivery room has received considerable attention in the last decade. In a paper reviewing the available literature on this subject, Richmond and Goldsmith³⁸ found that in both animal and human studies, although the results were mixed, there was a trend toward resuscitation with room air being as effective as using 100% oxygen. Animal studies showed that using air was nearly as effective as F_IO2 of 1.0 in reducing pulmonary vascular resistance and may prevent rebound pulmonary vascular resistance increases post-resuscitation. Most of the human studies only examined short-term outcomes, such as survival, Apgar score, and time to first breath, and most were not randomized.

The most recently published study, from the Benefits of Oxygen Saturation Targeting (BOOST) II Collaborative Group,³⁹ showed that in 3 RCTs including 2,448 extremely pre-term infants (< 28 weeks gestation), targeting oxygen saturation < 90% resulted in a statistically significant increased risk of death (P = .002), compared to the comparative group targeting saturation of 91–95%. Although the lower targeted saturation group had a significantly reduced incidence of ROP, the infants also had a significant increase in the rate of developing necrotizing enterocolitis.

Trauma

Patients suffering from multiple traumatic injuries are nearly always placed on supplemental oxygen, even if not intubated. Oxygen use in emergency care has been mandated in Advanced Trauma Life Support,⁴¹ Prehospital Trauma Life Support,⁴² and Advanced Cardiac Life Support,⁴³ despite scant evidence regarding efficacy and/or safety in this patient population. Oxygen supplementation begins at the point of injury and continues until presentation to the emergency department, usually via a non-rebreathing mask at 15 L/min, often despite S_pO2 readings of 100%. It has been witnessed on numerous occasions in our facility's emergency department: a patient brought in by the life squad wearing a non-rebreathing mask while talking on a cell phone, with the mouthpiece tucked under the mask. Clearly, oxygen administration was not indicated in this situation. The Prehospital Trauma Life Support guidelines⁴² for oxygen administration are based on the patient's spontaneous breathing frequency (Table 1). Other than with a normal breathing frequency,^12–20 the recommendation is to administer an F_IO2 of at least 0.85, with no mention of arterial oxygenation parameters.

Much of what the civilian medical community has learned about treating trauma victims was born from the military experiences in treating the casualties of war. Few studies have been conducted to determine how much oxygen trauma patients require, and most are observational. Stockinger and McSwain⁴⁴ retrospectively reviewed data from 5,090 spontaneously breathing trauma patients who presented to a civilian trauma center, in an attempt to determine the oxygen needs of trauma patients, in order to advance knowledge for military needs. Forty-three percent of the patients received oxygen, and they died more often than those who did not receive oxygen (2.3% vs 1.1%). Even after correcting for Injury Severity Score, mechanism of injury, and age, those who did not receive oxygen had no worse outcome than those who received oxygen, suggesting that supplementing oxygen does not improve outcomes in trauma patients who are not in respiratory distress.

Barnes et al conducted a prospective study to determine oxygen requirements and usage during transcontinental flights transporting mechanically ventilated wounded war fighters from Iraq to Germany with the Air Force Critical Care Air Transport Teams.⁴⁵ During the 6–8 hour flight an integrated computer recorded the ventilator settings and pulse oximetry readings. Oxygen was titrated according to the standard of care guidelines, keeping S_pO2 ≥ 94%. Twenty-two patients' data were recorded, resulting in 117 hours of continuous data. After calculating oxygen usage (L/min), it was found that the mean usage was 3.24 L/min, with a mean F_IO2 of 0.49 for all patients. Sixty-eight percent of the patients required ≤ 3 L/min, suggesting that oxygen requirements for trauma patients may be much lower than what is currently being administered.

Hemorrhagic shock is the leading cause of death in trauma patients, and poses a different set of problems when trying to determine appropriate oxygen administration. Hemorrhagic shock is a result of blood loss that may lead to decreased oxygen supply and cellular hypoxia despite normal arterial oxygenation indicators. Knight et al performed a literature review to increase understanding of the effects of oxygen administration following hemorrhagic shock.⁴⁶ The review found that F_IO2 of 1.0 and resuscitation are the most common treatments following hemorrhagic shock, although there is concern that hyperoxia may increase free radical formation and further cell damage. The literature suggested serum lactate should be monitored to assess cellular hypoxia and possibly guide oxygen administration, although the appropriate level remains unclear.

Despite anecdotal and sparse research data, there is no consensus for determining in which trauma patients to administer oxygen, and how much. The United States Special Operations Command's Tactical Combat Casualty Care guidelines state that oxygen may be beneficial for the following patients⁴⁷:

• Low oxygen saturation by pulse oximetry

• Injuries associated with impaired oxygenation

• Unconscious casualty

• Casualty in shock

• Casualty at altitude

• Casualty with traumatic brain injury (maintain S_pO2 > 90%)

Traumatic brain injury represents challenges when caring for a trauma patient, and is a leading cause of death and disability.⁴⁸ Much of the available literature's focus is on prehospital management of traumatic brain injury. It is well known that secondary brain injury can develop as a result of several factors, including inappropriate ventilation, glycemic control, cerebral edema, hypotension, and cerebral hypoxia.^49–51 Hypoxia has been identified as an independent risk factor for poor outcome with traumatic brain injury.⁵⁰ Providing oxygen to the injured brain is crucial to mitigating secondary brain injury, but the appropriate level of P_aO2 remains unclear, since adequate arterial oxygenation may not always equate to adequate brain oxygenation. Chi et al performed a prospective cohort study in 150 trauma patients with suspected head injury undergoing helicopter transport.⁵² The study goal was to determine the incidence of hypoxia and hypotension and to assess mortality and disability. Thirty-seven subjects had hypoxic episodes. The mortality for subjects without any secondary insults was 20%, versus 37% for those who had hypoxic episodes. Surviving subjects who experienced hypoxia also had a greater degree of disability at hospital discharge. In an attempt to determine the relationship between hypoxemia and hyperoxemia and outcome, Davis et al performed a retrospective review of 3,420 subjects treated for traumatic brain injury.⁵³ The study found that mild hyperoxemia (P_aO2 110–487 mm Hg) was associated with increased survival, while hypoxemia (P_aO2 < 110 mm Hg) and extreme hyperoxemia (P_aO2 > 487 mm Hg) were associated with increased mortality.

Although monitoring and treating intracranial pressure remains the standard of care, devices to measure brain-tissue oxygenation are being utilized. Martini and associates⁵⁴ reviewed the available published literature on this practice and found that monitoring brain-tissue oxygenation has shown value in determining poor prognosis following traumatic brain injury, and that interventions to increase cerebral perfusion pressure and P_aO2 can result in increased brain-tissue oxygenation. The authors' review also found that retrospective studies suggest that maintaining a target brain-tissue oxygenation (usually 20 mm Hg) may have potential benefits, but prospective studies showed no outcome benefits.

ARDS

ARDS was first described by Ashbaugh et al in 1967. Historically, the mortality rate for ARDS was reportedly 40–60%^55–58 by most accounts, until the turn of the 21st century, when new therapeutic studies emerged that improved outcome. The ARDS Network clinical trial, published in 2001, was the first evidence that ARDS mortality can be improved by changes in mechanical ventilation practice.⁵⁹ This landmark study showed that reducing tidal volumes to as low as 4–6 mL/kg of ideal body weight reduced mortality by 22%. Additionally the study supported oxygenation by the use of a PEEP/F_IO2 table to maximize lung recruitment and minimize oxygen exposure, due to earlier evidence in animal models that high F_IO2 may be toxic. The targeted range for oxygenation was P_aO2 55–88 mm Hg and S_pO2 88–95%. Although the best strategy for using PEEP and F_IO2 has not been identified, mounting evidence suggests the use of the lowest F_IO2 possible and the use of adequate PEEP to increase oxygenation without producing cardiovascular side effects.⁶⁰

Kallet and Branson⁶¹ performed a literature review in an effort to determine if the ARDS Network study's PEEP/F_IO2 table is the best method for maintaining oxygenation and minimizing oxygen exposure. The authors found that, since the PEEP required for most patients with ARDS is relatively low, the use of the ARDS Network PEEP/F_IO2 table is supported by high level evidence, although there is a small subset of patients who may require an individualized approach to setting PEEP and F_IO2.

Myocardial Infarction

According to Centers for Disease Control statistics, heart disease is the leading cause of death in the United States, accounting for more than 600,000 fatalities annually.⁶² Myocardial infarction accounts for more than half of these deaths. For more than 100 years oxygen has been used to treat myocardial infarction and angina,⁶³ with little evidence as to the efficacy or potential harm of this practice. Oxygen administration can cause vasoconstriction, regardless of arterial saturation, and raise blood pressure and lower cardiac oxygen consumption, heart rate, and cardiac index.^64–66 Foster et al found that as P_aO2 increased, so did arterial pressure and systemic vascular resistance.⁶⁷ Kenmure et al⁶⁸ and Thomas et al⁶⁹ found an increase in blood pressure and decrease in cardiac output when patients suffering from a myocardial infarction breathed F_IO2 of 0.4. McNulty et al, using a Doppler flow wire, showed in 18 subjects that coronary vascular resistance increased by 41% and coronary blood flow decreased by 29% when the subjects breathed F_IO2 of 1.0 for 15 min.⁷⁰ These works showed the physiologic effects of oxygen administration on the coronary system, but the effect on outcome was not evaluated.

Wijesinghe et al performed a review of the published literature that included RCTs of oxygen therapy in myocardial infarction.⁷¹ Of 51 potential studies, only 2 met the inclusion criteria. One of the studies of 200 subjects randomized to either room air or 6 L/min oxygen for 24 hours after having a myocardial infarction found that deaths and the incidence of ventricular tachycardia were higher in the oxygen group, but the difference was not statistically significant. Opiate use was not different between the groups. The other study randomized 50 subjects to either room air or 4 L/min oxygen for 24 hours. Although more subjects experienced an episode of oxygen desaturation, 80% in the room air group (P < .01), there was statistically no difference in the incidence of ventricular tachycardia and opiate use between groups. Mortality was not evaluated.

Kones's review of oxygen use for acute myocardial infarction found there are no large randomized studies available for evaluation.⁷² He found that the evidence supporting oxygen use in patients having acute myocardial infarction but who had normal oxygen saturation was old and of poor quality. Kones noted that recent physiological evidence that oxygen use in this patient population that are not hypoxemic suggests that there is no evidence of benefit and may be harmful. His conclusion was that, in these patients, oxygen should be administered only if saturation drops below 94%, although there is no evidence to support this recommended saturation level.

A recent Cochrane Collaborative meta-analysis cited 3 RCTs comparing groups given oxygen or air when experiencing a myocardial infarction.⁷³ The 3 studies included 387 subjects, with 14 of those dying. Of those 14, nearly 3 times as many subjects in the oxygen group died, compared to those given air. Although this suggests that oxygen administration may be harmful, definitive conclusions cannot be drawn because the studies had small numbers of subjects, so the results may have happened by chance. The authors' conclusion was that a large RCT is required to refute or confirm these findings.

Cardiac Arrest

Cardiac arrest often results from a myocardial infarction. Even if return of spontaneous circulation is achieved, nearly 60% of these patients will not survive.⁷⁴ The high mortality has been associated with anoxic brain injury, cardiac stunning, and reperfusion injury.⁷⁵ High concentration oxygen administration during the post-cardiac-arrest period has been questioned as a potential contributor to the high mortality after return of spontaneous circulation. Kilgannon and associates conducted 2 multicenter cohort studies^76,77 using the Project IMPACT critical care database to examine the effect of hyperoxia after cardiac arrest and the effect on mortality. The first study⁷⁶ included 6,326 subjects, and the end point was in-hospital mortality. The subjects were divided into 3 groups: hyperoxia (defined as P_aO2 ≥ 300 mm Hg), hypoxia (defined as P_aO2 < 60 mm Hg or P_aO2/F_IO2 < 300 mm Hg), and normoxia (defined as P_aO2 60–300 mm Hg).

Of the 6,326 subjects, 18% had hyperoxia, 63% had hypoxia, and 19% had normoxia. The hyperoxia group had significantly higher in-hospital mortality (63%) than did the normoxia group (45%) or the hypoxia group (57%). In the second study⁷⁷ using the Project IMPACT database, Kilgannon's group evaluated 4,459 subjects post-cardiac-arrest to determine the relationship between P_aO2 and in-hospital mortality. Of the 4,459 subjects, 54% died. The observed P_aO2 values were divided into 5 groups: 60–99, 100–199, 200–299, 300–399, and ≥ 400 mm Hg. The results of the study showed that there was an association between increased P_aO2 and increased mortality, even in those subjects who did not have supranormal P_aO2. For every 100 mm Hg increase in P_aO2 there was a 24% increase in the relative risk of death. Interestingly, a 25% increase in P_aO2 resulted in a 6% relative risk of death. The results of this study suggest that since a relatively small increase in P_aO2 increases mortality, limiting supplemental oxygen as much as possible after cardiac arrest may be beneficial. It is hypothesized that increased free radical formation caused by high concentration delivery, along with reperfusion injury, may be responsible for the increased mortality.

Congestive Heart Failure

Patients with congestive heart failure often suffer from dyspnea and hypoxia. High concentration oxygen is often given to these patients, despite previous studies showing that administering F_IO2 of 1.0 to healthy subjects decreases cardiac output and increases systemic vascular resistance.^78,79 Little is known about the hemodynamic effects of oxygen administration in these patients. Haque et al conducted a small study in which 22 subjects with class 3 and 4 heart failure were divided into 3 separate experiments.⁸⁰ Experiment 1 involved 10 subjects having hemodynamic variables measured while breathing room air and then after breathing an F_IO2 of 1.0 for 20 min. Experiment 2 involved 7 subjects having the same hemodynamic measurements collected after breathing room air and then after 5 min on F_IO2 of 0.24, 0.40, and 1.0. F_IO2 of 1.0 significantly reduced cardiac output and stroke volume, and increased pulmonary capillary wedge pressure and systemic vascular resistance, as compared to breathing room air (P < .01). Graded oxygen showed a progressive decrease in cardiac output (P < .001) and stroke volume (P < .02), and an increase in systemic vascular resistance (P < .005). Additionally, S_aO2 progressively increased, from 93.6 ± 1.5% on room air to 100.0 ± 0% on F_IO2 of 1.0. Based on the results of this small study, the authors recommend that, in the absence of hypoxemia, oxygen should be used cautiously with patients suffering from severe congestive heart failure.

Stroke

Oxygen is frequently administered to patients suffering from a stroke in the prehospital setting, and is often continued in the hospital, despite current guidelines that recommend not administering oxygen to non-hypoxic patients.⁸¹ The pervasive idea that oxygen therapy is beneficial stems from the fact that ischemic stroke causes a decrease in oxygen to the brain, resulting in tissue hypoxia and cell death. The prevailing logic is that neuroprotection can be achieved by raising oxygen levels in ischemic tissues.⁸² Extending the logic further, it was thought that hyperbaric oxygen therapy, which can produce extreme hyperoxemia, would be beneficial for stroke patients, but clinical trials failed to show any benefit.^83–85 It is well established that hyperoxemia increases free radical formation and could induce cerebral vasoconstriction and reduced blood flow.^79,86 Animal studies have shown increased mortality when exposed to high oxygen levels following cerebral ischemia.^87,88

Pancioli and associates performed a retrospective chart review of 167 non-intubated, ischemic stroke patients totaling 600 in-patient days at a university hospital to determine whether these patients had indications for supplemental oxygen.⁸⁹ The criteria used for supplemental oxygen therapy are listed in Table 2. Sixty-one percent of the subjects received supplemental oxygen at some point during their hospital stay, which accounted for 322 days of receiving oxygen. Of those 322 days, 46% met at least one of the pre-established criteria for oxygen use. Of the 348 days in which criteria for supplemental oxygen were not met, the subjects still received oxygen 46% of the time. The authors estimated that not giving oxygen when it is not indicated could produce up to 45% savings in resources. Ronning and Guldvog⁹⁰ conducted an RCT including 500 subjects to determine whether F_IO2 of 1.0 for the first 24 hours after stroke would reduce mortality, neurological impairment, or disability, as compared to receiving no oxygen. The subjects in the room air group had a higher 1 year survival, but the difference was not statistically significant (P = .30). For subjects with severe stroke there was a statistically nonsignificant tendency toward a higher 1 year survival in the oxygen group (P = .60). Neurological impairment and disability did not differ between the 2 groups. The authors concluded that oxygen should not routinely be given to patients suffering from acute stroke.

Wound Infection

Surgical wound infection is a serious complication that can increase hospital stays and costs,^91–93 and increase morbidity and mortality.^94,95 Bacterial tissue contamination establishes wound infections within a few hours post-surgery,⁹⁶ so interventions during this time have the greatest potential to prevent a severe infection. Prophylactic antibiotic therapy is the most common perioperative intervention to prevent wound infection. Due to laboratory evidence that oxidative bactericidal activity is highly dependent on increasing the oxygen tension in a wound,⁹⁷ it has been suggested that providing high levels of perioperative oxygen may attenuate bacterial wound infections.

Greif and associates⁹⁸ conducted an RCT including patients undergoing colorectal surgery to receive F_IO2 of either 0.3 or 0.8 intraoperatively, and for 2 hours postoperatively. All subjects received prophylactic antibiotic therapy. Wounds that were culture positive were considered infected. Subjects in the F_IO2 0.8 group had significantly less wound infections, versus those in the F_IO2 0.3 group (5% vs 11%, P = .01). Hospital lengths of stay were similar.

In a smaller study, conducted in Israel, 38 subjects undergoing elective colorectal surgery were also randomized to receive the same oxygen concentrations and length of therapy as in the Greif study.⁹⁹ The wound infection rate in the F_IO2 0.8 group was higher than in the F_IO2 0.3 group, but the difference was not statistically significant (P = .53), although this could have been due to the small sample size. Even though the infection rates were not lower in the high oxygen group, the authors could not make a definitive recommendation for the use or non-use of high oxygen concentration.

Postoperative Nausea and Vomiting

Postoperative nausea and vomiting (PONV) is common, with an occurrence of 20–70% despite current pharmaceutical interventions.^100–102 The unpleasantness for the patient notwithstanding, PONV can increase the risk of aspiration pneumonia and can lead to delayed discharge and unexpected hospital admissions following surgery.¹⁰³ Recent research suggests that supplemental oxygen may have a positive effect on PONV following selected surgical procedures.

Greif et al conducted an RCT in 231 subjects undergoing colon resection, to receive F_IO2 of either 0.8 or 0.3 during surgery and 2 hours afterward.¹⁰⁴ The incidence of PONV during the first 24 hours postoperatively was recorded. PONV was observed in 17% of the subjects who received F_IO2 0.8, versus 30% in the F_IO2 0.3 group (P = .03).

Ghods et al¹⁰⁵ randomized 106 subjects undergoing cesarean birth to receive 8 L/min oxygen for 6 hours postoperatively or 5 L/min in the recovery room and no oxygen thereafter, and evaluated the incidence of PONV during the first 6 postoperative hours. PONV occurred in 28% of subjects in the 8 L/min group, and nearly 25% in the control group. The difference between groups was not statistically significant (P = .66).

Joris et al¹⁰⁶ conducted an RCT randomizing 150 subjects to receive either F_IO2 of 0.3, F_IO2 of 0.8, or F_IO2 of 0.3 oxygen with droperidol, during thyroidectomy, and evaluated the incidence of PONV for 24 hours post-surgery. There was no difference in the incidence of PONV in the F_IO2 0.3 and 0.8 groups (48% vs 46%), but the group receiving F_IO2 of 0.3 plus droperidol was 22%, which was statistically different from the other 2 groups (P = .004). Time to first meal was significantly shorter in the droperidol group.

Treschan et al¹⁰⁷ randomly assigned 210 subjects having strabismus surgery to the same study arms as the Joris study, with the difference being the use of ondansetron instead of droperidol. PONV was evaluated postoperatively at 6 and 24 hours. As opposed to the Joris study, there was no statistical difference in the incidence of PONV between any of the 3 groups (P = .28), although the incidence was lower for the ondansetron group (28%) versus the F_IO2 0.8 group (38%) and the F_IO2 0.3 group (41%). The low number of subjects in each group may account for the lack of statistical significance.

Cluster Headache

The first description of cluster headaches (CH) was given by London neurologist Wilfred Harris in 1926.¹⁰⁹ His treatment for these subjects was alcohol injections around the supraorbital and infraorbital nerve. Horton first described the use of oxygen for the treatment of CH in 1952,¹¹⁰ and was brought to the forefront by Kudrow in 1981.¹¹¹ This first systematic study compared oxygen by mask at 7 L/min for 15 min versus sublingual ergotamine. The study showed that both treatments were effective in aborting CH attacks, but oxygen aborted over 70% of the attacks in 82% of the subjects, whereas ergotamine worked as well in only 70% of the subjects. The average response time with oxygen to abort the CH was 6 min, versus 10–12 min with ergotamine.

In 1985, Fogan conducted a small double-blind crossover study comparing oxygen versus air, both at 6 L/min, for the treatment of CH.¹¹² Subjects scored their degree of relief with each therapy, with the relief score being significantly higher when inhaling oxygen versus air (P = .01). The average relief score was 1.93 for oxygen inhalation and 0.77 for air inhalation, out of a possible 3.

More recently, Garza conducted a double-blind RCT of 109 subjects with CH to alternately receive 12 L/min oxygen or air via mask for 15 min at the onset of an attack.¹¹³ The primary end point was complete or adequate pain relief at 15 min. The results showed that the primary end point was reached 78% of the time with oxygen inhalation, versus 20% with air (P = .001). Oxygen was also superior to air concerning the secondary end points of pain free at 30 min, pain reduction at 60 min, and need for additional medication 15 min after treatment.

Although the efficacy of oxygen administration for treatment of CH is well documented, there have been observations of rebound CH post-treatment. The rebound effect is defined as a CH that returns more rapidly than usual following complete relief after oxygen inhalation, or an increased number of attacks in a 24 hour period. Geerlings et al performed a retrospective study and found 8 subjects who experienced rebound CH.¹¹⁴ In these subjects the mean duration until the next CH was 39 min after using oxygen to treat the previous CH versus 933 min if oxygen was not used. The mean frequency of CH per day was 4.1 when using oxygen, versus 2.5 without using oxygen. It is hypothesized that use of lower flow (7 L/min or less) may lead to rebound CH in susceptible patients. Cohen et al evaluated the effectiveness of using 12–15 L/min oxygen for treatment of CH.¹¹⁵ While headache relief was comparable to studies using lower flows, no rebound CH were reported.

Carbon Monoxide Poisoning

Carbon monoxide (CO) poisoning is the leading cause of poisoning death in the United States. CO poisoning is sometimes overlooked because the clinical signs and symptoms are not the same for all patients. It is well established that hemoglobin's affinity for CO is over 200 times higher than for oxygen and that blood CO levels in excess of 20% can affect the brain and heart, due to their high metabolic rate.¹¹⁶ Tissue hypoxia is the hallmark of CO poisoning, so oxygen is the standard treatment, although pulse oximetry readings are unreliable due to the device being unable to distinguish between oxygen and CO bound to the hemoglobin. In the late 1800s, Haldane showed that high oxygen tension can counteract the hemoglobin to CO affinity.¹¹⁷ The half-life of CO while breathing room air is approximately 5 hours. Breathing normobaric F_IO2 of 1.0 reduces the half-life to 1 hour, and hyperbaric oxygen therapy reduces the half-life to 20 min.^118–120

Breathlessness

One of the most controversial and misunderstood uses of supplemental oxygen is for patients experiencing breathlessness. Breathlessness is a common symptom of advanced lung, cardiac, and neuromuscular disease, and the intensity increases as death approaches.^121,122 Even with increased understanding of breathlessness and the pharmacologic and non-pharmacological interventions available, it remains difficult to manage. Breathlessness makes caregivers and healthcare providers feel helpless, further complicating management. Upon conducting a survey, Abernethy and associates found that 70% of clinicians would prescribe oxygen for breathlessness despite normal oxygen saturation, and 35% would prescribe oxygen if the patient asked for it.¹²³ Hypoxemia does not appear to be the driving force in chronic breathlessness.

Abernethy et al conducted a double-blind RCT in 239 subjects with refractory breathlessness, and evaluated the effectiveness of administering 2 L/min oxygen, as compared to 2 L/min air.¹²⁴ The study results showed that morning breathlessness improved more in the oxygen group, but improved more in the evening with the air group. Improvement in quality of life was no different between groups, nor was there a difference in breathlessness over a 24 hour period. Breathlessness scores of subjects with moderate to severe breathlessness improved most, irrespective of the treatment arm. The authors concluded that the study results suggest it is the flow of gas through the nasal passages that improves the feeling of breathlessness, regardless of whether oxygen or air is used.

Johnson et al conducted a meta-analysis of the available literature focusing on oxygen to treat chronic refractory breathlessness.¹²⁵ Of the 13 studies reviewed, 2 showed a benefit when supplementing oxygen to breathless subjects. These 2 studies involved COPD patients, and the benefit of oxygen administration was small and limited to breathlessness as a result of exertional desaturation in one study. The remaining studies show no benefit of administering oxygen as opposed to air.

What the Literature Says

Oxygen is administered for many diseases and conditions in hospitalized patients. The evidence in the literature suggests that supplemental oxygen is clearly indicated in the following instances: reversal of hypoxemia, traumatic brain injury, hemorrhagic shock, resuscitation during cardiac arrest, and CO poisoning.

Oxygen should be administered to target an S_pO2 of 94–98%, except with CO poisoning, due to the inaccuracy of pulse oximetry. Patients with COPD, neuromuscular disease, and obesity who are at risk for hypercapnia should have a target S_pO2 of 88–92%. Patients with ARDS should have a target S_pO2 of 88–95%, due to evidence from the ARDS Network trial. Infants should have a target S_pO2 of 88–94%, depending on gestational age, to prevent ROP, bronchopulmonary dysplasia, and cerebral palsy.

Summary

Oxygen is a popular drug and is often administered indiscriminately. The belief that oxygen is harmless and the attitude of “if a little is good, more is better” is common in today's healthcare environment. Severinghaus and Astrup proclaimed that “If introduced today, this gas might have difficulty getting approved by the Food and Drug Administration.”¹²⁶ Priestley's words may be even truer today: “The air which nature has provided for us is as good as we deserve.”¹