association between hyperoxia and mortality after stroke- a multicenter cohort study.pdf

Neurologic Critical Care

Critical Care Medicine www.ccmjournal.org 387

Objective: To test the hypothesis that hyperoxia was associated with higher in-hospital mortality in ventilated stroke patients admit-ted to the ICU.Design: Retrospective multicenter cohort study.Setting: Primary admissions of ventilated stroke patients with acute ischemic stroke, subarachnoid hemorrhage, and intracere-bral hemorrhage who had arterial blood gases within 24 hours of admission to the ICU at 84 U.S. ICUs between 2003 and 2008. Patients were divided into three exposure groups: hyperoxia was defined as Pao2 ≥300 mm Hg (39.99 kPa), hypoxia as any Pao2<60 mm Hg (7.99 kPa) or Pao2/Fio2 ratio ≤300, and nor-

moxia, not defined as hyperoxia or hypoxia. The primary outcome was in-hospital mortality.Participants: Two thousand eight hundred ninety-four patients.Methods: Patients were divided into three exposure groups: hyper-oxia was defined as Pao2 more than or equal to 300 mm Hg (39.99 kPa), hypoxia as any Pao2 less than 60 mm Hg (7.99 kPa) or Pao2/Fio2 ratio less than or equal to 300, and normoxia, not defined as hyperoxia or hypoxia. The primary outcome was in-hospital mortality.Interventions: Exposure to hyperoxia.Results: Over the 5-year period, we identified 554 ventilated patients with acute ischemic stroke (19%), 936 ventilated patients with subarachnoid hemorrhage (32%), and 1,404 ventilated patients with intracerebral hemorrhage (49%) of whom 1,084 (38%) were normoxic, 1,316 (46%) were hypoxic, and 450 (16%) were hyperoxic. Mortality was higher in the hyperoxia group as compared with normoxia (crude odds ratio 1.7 [95% CI 1.3-2.1]; p < 0.0001) and hypoxia groups (crude odds ratio, 1.3 [95% CI, 1.1–1.7]; p < 0.01). In a multivariable analysis adjusted for admis-sion diagnosis, other potential confounders, the probability of being exposed to hyperoxia, and hospital-specific effects, expo-sure to hyperoxia was independently associated with in-hospital mortality (adjusted odds ratio, 1.2 [95% CI, 1.04–1.5]).Conclusion: In ventilated stroke patients admitted to the ICU, arterial hyperoxia was independently associated with in-hospital death as compared with either normoxia or hypoxia. These data underscore the need for studies of controlled reoxygenation in ventilated critically ill stroke populations. In the absence of results from clinical trials, unnecessary oxygen delivery should be avoided in ventilated stroke patients. (Crit Care Med 2014; 42:387–396)Key Words: brain injury; intracerebral hemorrhage; ischemic stroke; resuscitation; subarachnoid hemorrhage

Resuscitative efforts in the emergency department (ED) or ICU after stroke include aggressive airway management and ventilation with supplemental oxygen in those patients

unable to maintain a normal respiratory function (1). The classic paradox with oxygen supplementation is that although oxygen is

Copyright © 2013 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins

DOI: 10.1097/CCM.0b013e3182a27732

*See also p. 469.1Department of Neurology, Thomas Jefferson University, Philadelphia, PA.2Division of Critical Care and Neurotrauma, Department of Neurosurgery, Thomas Jefferson University, Philadelphia, PA.

3Division of Biostatistics, Rothman Institute of Thomas Jefferson Univer-sity, Philadelphia, PA.

4Department of Medicine, Thomas Jefferson University, Philadelphia, PA.5Division of Cerebrovascular Diseases, Thomas Jefferson University, Philadelphia, PA.

Drs. Rincon, Kang, and Maltenfort had full access to all of the data in the study and take responsibility for the integrity of the data and the accu-racy of the data analysis. They conceived and designed the study and analyzed and interpreted. Drs. Rincon and Maltenfort acquired the data and performed statistical analysis. Drs. Rincon, Kang, Maltenfort, Vibbert, Athar, Jallo, McBride, and Bell drafted the manuscript for important intel-lectual content. Drs. Rincon, Jallo, and Bell gave administrative, technical, or material support and supervised the study.

Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/ccmjournal).

Dr. Rincon has received salary support from the American Heart Asso-ciation (AHA 12CRP12050342). He received support for travel from AAN, NCS, SCCM, and IntouchHealth. His institution received grant sup-port from the AHA. Dr. Jallo’s institution received grant support from the Department of Defense (#110398). Neither funding sources nor Cerner Corporation had a role in the design of this study or in the decision to submit it for publication. The remaining authors have disclosed that they do not have any potential conflicts of interest.

For information regarding this article, E-mail: [email protected]

Association Between Hyperoxia and Mortality After Stroke: A Multicenter Cohort Study*

Fred Rincon, MD, MSc, MBE, FACP, FCCP, FCCM1,2; Joon Kang, MD1; Mitchell Maltenfort, PhD3;

Matthew Vibbert, MD1,2; Jacqueline Urtecho, MD1,2; M. Kamran Athar, MD2,4;

Jack Jallo, MD, PhD, FACS2; Carissa C. Pineda, MD1,5; Diana Tzeng, MD1,5;

William McBride, MD1,5; Rodney Bell, MD, FAHA1,2,5

http://journals.lww.com/ccmjournal

mailto:[email protected]

Rincon et al

388 www.ccmjournal.org February 2014 • Volume 42 • Number 2

necessary to maintain energy metabolism, excessive oxygen may potentiate primary or secondary brain injury by triggering free radical production, activation of apoptotic cascades, and facili-tating organ-specific hypoperfusion (2–9).

Because tissue hypoxia plays a critical role in primary and secondary neuronal damage, normobaric hyperoxia is generally thought a logical stroke therapy despite controversial results of animal models of ischemic (10–15) and hemorrhagic stroke (16) and clinical studies (8, 17–19). Although animal data from stroke models demonstrate that hyperoxia is associated with increased oxidative stress, worst ischemia-induced brain dam-age, and overall mortality (10–12), clinical and observational data in critically ill stroke populations are currently lacking.

To this end, the objective of this study was to define the rela-tionship between hyperoxia and outcome in ventilated stroke patients. Specifically, we wanted to determine: 1) the occur-rence of hyperoxia at admission to the ICU and 2) whether hyperoxia at admission to the ICU was an early predictor of in-hospital mortality after adjustment for other confound-ers in a multivariable analysis using a robust multicenter ICU repository. We hypothesized that hyperoxia in ventilated stroke patients would be associated with higher in-hospital mortality.

METHODS

Study Design and Patient PopulationIt is a retrospective multicenter cohort study using a pro-spectively compiled and maintained robust registry (Cerner

Corporation—Project IMPACT [PI], Bel Air, MD). PI is a large administrative database designed for critical care units across all disciplines. Adult ICUs from 131 U.S. hospitals par-ticipated in PI, and data from more than 400,000 patients were collected prospectively. All patients admitted to a partici-pating ICU were included regardless of diagnosis or outcome. Participating institutions uploaded data quarterly to a central repository. Data fields included hospital and ICU organiza-tional characteristics, demographics, comorbid conditions, physiologic data, procedures, in-hospital complications, hos-pital and ICU length of stay (LOS), and discharge status. Data were collected by dedicated onsite personnel trained and cer-tified by PI. Personnel also required passing a written certi-fication examination to ensure uniformity in both database definitions and entry. The ICUs in PI repository represent a wide scope of practice environments, including medical, sur-gical, neurologic, and multidisciplinary ICUs. The institutions are heterogeneous in terms of hospital size, type (community vs academic; private vs public), and location (urban, subur-ban, or rural). Participating hospitals are not restricted to any particular geographic region.

For this analysis, ventilated patients with a primary diagno-sis of acute ischemic stroke (AIS), aneurysmal subarachnoid hemorrhage (SAH), and intracerebral hemorrhage (ICH), older than 17 years, and consecutively admitted to an ICU from 2003 to 2008 were selected (Fig. 1). The following vari-ables were recorded for all patients admitted to the ICU: ED boarder status (20), demographic variables, comorbidities,

most abnormal physiologic variables, first arterial blood gas (ABG) over the first 24 hours in the ICU, Simplified Acute Physiology Score II score (21), and admission and hospital discharge status (alive independent, alive partially dependent, alive fully depen-dent, or dead) (22).

To assess for the effect of medical complications on in-hospital mortality, we collected data on any of the following organ dysfunc-tions during the ICU stay according to the definitions of PI: cardiovascular (sys-tolic blood pressure [SBP] < 90 mm Hg, or mean arterial pressure [MAP] < 70 mm Hg, or vasopressor requirement to keep SBP > 90 mm Hg or MAP > 70 mm Hg, and duration > 1 hr), metabolic (lactic aci-dosis > 2.0 mmol/dL), respi-ratory (acute lung injury Pao

2/Fio

2 less than or equal to

Figure 1. Description of study cohort. ABG = arterial blood gas, AIS = ischemic stroke, SAH = subarachnoid hemorrhage, ICH = intracerebral hemorrhage.



TablE 1. Demographic Characteristics of the Cohort of Ventilated Stroke Patients

Characteristics

With arterial blood Gases

pTotal (n = 2,894) Hypoxia (n = 1,084) Normoxia (n = 1,316) Hyperoxia (n = 450)

Age, mean (sd), yr 61 (15) 61 (15) 60 (17) 61 (15) NS

Female sex, n (%) 1,408 (49) 715 (51) 469 (43) 203 (45) < 0.0001

Race/ethnicity (n = 2,763), n (%) < 0.0001

White 2,033 (74) 980 (78) 718 (69) 311 (72)

Black 477 (17) 186 (14) 203 (20) 75 (17)

Latino/Hispanic 155 (6) 50 (5) 71 (6) 34 (5)

Asian/Pacific Islander 74 (3) 14 (1) 43 (4) 17 (2)

Native American 12 (< 1) 3 (< 1) 5 (< 1) 4 (< 1)

Preadmission functional statusa (n = 2,807) NS

Independent 2,285 (81) 1043 (81) 852 (81) 363 (84)

Partially dependent 268 (10) 120 (10) 100 (9) 41 (9)

Fully dependent 252 (9) 119 (9) 96 (9) 30 (7)

DNR status, n (%) NS

Full code 2,718 (95) 1232 (5) 1017 (5) 428 (5)

DNR/no cardiopulmonary resuscitation/limited

intervention

137 (5) 49 (5) 58 (5) 19 (5)

Emergency department boarder 112 (3) 47 (5) 49 (4) 16 (2) 0.02

Hospital size, n (%) 0.002

Extralarge 1,446 (50) 638 (48) 602 (52) 191 (42)

Large 1,231 (43) 580 (40) 478 (41) 216 (48)

Small to medium 217 (7) 98 (7) 73 (6) 43 (10)

Hospital location, n (%) NS

Urban 1,907 (66) 850 (65) 779 (68) 289 (64)

Suburban 516 (18) 229 (17) 207 (18) 99 (22)

Rural 471 (16) 173 (17) 167 (14) 131 (18)

Hospital type, n (%) 0.01

Academic 1,011 (35) 454 (35) 395 (34) 167 (37)

Community nonacademic 1,659 (57) 758 (58) 673 (58) 247 (55)

Public 224 (8) 104 (7) 85 (7) 36 (8)

Diagnosis, n (%) NS

AIS 554 (19) 268 (20) 211 (18) 75 (17)

SAH 936 (32) 401 (30) 380 (33) 135 (30)

ICH 1,404 (49) 647 (49) 499 (46) 240 (53)

Median (interquartile range) 6 (3–9) 6 (3–10) 6 (3–9) 5 (3–8) 0.001

Simplified Acute Physiologic Score II, mean (sd)

48 (15) 50 (16) 45 (14) 51 (16) < 0.0001

Comorbidities,b n (%)

Cardiovascular disease 113 (4) 45 (4) 39 (3) 29 (4) NS

(Continued)

Rincon et al


300 [39.99 kPa] or positive end-expiratory pressure > 5 cm H

2O), renal (serum creatinine increased > 1 mg/dL from

baseline despite fluid resuscitation or creatinine > 2.0 mg/dL regardless of baseline), hepatic (acute elevation of serum bilirubin > 2 mg/dL), hematologic (platelet count < 100,000/mm3 or prothrombin time/thromboplastin time > 1.5 times baseline), and neurologic (onset delirium or Glasgow Coma Scale [GCS] < 12), and need for intracranial pressure (ICP) monitoring. These definitions have been used in prior studies from the PI repository (23, 24).

Hospitals were divided by hospital location into urban, suburban, and rural; by type into community (nonacademic), academic (university-based), and public; and according to the Halpern criteria (25) into small to medium size (< 300 beds), large (301–499 beds), and extra large (> 500 beds). We also collected the total hospital LOS. Specific data on the National Institutes of Health Stroke Scale, the Hunt and Hess Score,

the ICH score, or other index of disease severity or admission imaging were not available from the PI repository. Based on the de-identified nature of the database, the Institutional Review Board of Thomas Jefferson University Hospital (Philadelphia, PA) exempted this analysis from full review.

Definition of Main Exposure and Outcome VariablesPatients were divided into three exposure groups defined a priori based on the Pao

2 or ratio of Pao

2 to inspired Fio

2 on

the first ABG at admission to the ICU. Hyperoxia was defined as Pao

2 of 300 mm Hg (39.99 kPa) or greater (24, 26), hypoxia

as a Pao2 of less than 60 mm Hg (7.99 kPa) or Pao

2/Fio

2 ratio

less or equal than 300 (27, 28), and normoxia not classified as hyperoxia or hypoxia (24, 26). The primary outcome measure was in-hospital mortality. A secondary analysis compared the outcomes based on exposure groups according to neurologic diagnosis.

Pulmonary disease 68 (2) 39 (4) 21 (2) 5 (1) 0.02

Renal disease 65 (2) 28 (3) 20 (2) 17 (2) NS

Chronic liver disease 30 (1) 10 (1) 9 (< 1) 11 (2) NS

Chronic HIV infection 10 (< 1) 1 (< 1) 6(< 1) 3 (< 1) NS

Cancer 26 (< 1) 10 (1) 10 (< 1) 6 (< 1) NS

Organ dysfunction, n (%)

Cardiovascular 581 (20) 274 (21) 196 (18) 103 (23) < 0.0001

Metabolic (lactic acidosis) 95 (3) 46 (4) 28 (3) 18 (4) 0.004

Respiratory 744 (26) 457 (35) 190 (18) 89 (20) < 0.0001

Renal 159 (5) 98 (7) 46 (4) 19 (4) 0.002

Hematologic 164 (6) 54 (5) 61 (5) 49 (7) NS

Hepatic 59 (2) 18 (2) 28 (2) 13 (2) NS

Neurologic 116 (4) 54 (4) 42 (4) 18 (4) NS

Intracranial pressure monitor (n = 2,873), n (%)

994 (35) 442 (34) 388 (36) 149 (33) NS

AIS 40 (7) 22 (8) 15 (7) 3 (4)

SAH 508 (55) 216 (54) 212 (56) 72 (53)

ICH 446 (32) 204 (32) 161 (32) 74 (31)

NS = not significant, DNR = do not resuscitate, AIS = ischemic stroke, SAH = subarachnoid hemorrhage, ICH = intracerebral hemorrhage, GCS = Glasgow Coma Scale.aThe definition of functional status in the Project IMPACT database is independent (the patient is living at home requiring no assistance in completing activities of daily living, which includes people who are homeless, or who are incarcerated, but otherwise physically and mentally functional); partially dependent (the patient is living at home, in a group home or in a care facility, and requires some assistance in completing the activities of daily living, and the limitation(s) requiring assistance may be physical or mental); and fully dependent (the patient is living at home or in a care facility and is unable to perform the activities of daily living; must be cared for by other(s); the limitations requiring assistance may be physical or mental).bCardiovascular disease (defined as baseline symptoms such as angina or shortness of breath at rest or on minimal exertion, New York Heart Association class IV, plus one or more of the following diagnoses: severe coronary artery disease, severe valvular heart, or severe cardiomyopathy), respiratory disease (defined as chronic obstructive, restrictive, or vascular pulmonary disease resulting in severe exercise restriction, such as unable to climb stairs or perform household duties; or respirator dependency related to active respiratory disease; or documented chronic hypoxia, hypercapnia, or pulmonary hypertension > 40 mm Hg), cirrhosis, chronic renal disease or end-stage renal disease, HIV status, and cancer.Data presented as n (%), mean (sd), or median (interquartile range).

TablE 1. (Continued). Demographic Characteristics of the Cohort of Ventilated Stroke Patients

Characteristics

With arterial blood Gases

pTotal (n = 2,894) Hypoxia (n = 1,084) Normoxia (n = 1,316) Hyperoxia (n = 450)



Statistical analysesContinuous data are presented as mean and sd or median and interquartile rang as appropriate based on distribution of the data; categorical data are reported as proportions and 95% CIs. For evaluation of differences at the univariate level, the one-way analysis of variance test and the difference of means were compared with the Tukey-Kramer post hoc test, the Wil-coxon-Kruskal-Wallis test for nonnormally distributed data, or the chi-square test or Fisher exact test for proportions. The primary outcome for the cohort as a whole, and for subgroups based on neurologic diagnosis (AIS, SAH, or ICH), were com-pared between exposed groups using the chi-square test for the difference in proportions. We then calculated odds ratios (OR) and 95% CI, and to test for significance, we applied a Bonferroni correction for multiple pairwise comparisons. In this case, for three exposure groups, a significant α level of 0.017 was accepted (0.05 divided by 3 or 0.017). For days to primary outcome analysis, Kaplan-Meier survival estimates and log-rank tests were used to compare the exposure group of interest. For the multivariable analysis, generalized esti-mating equations (GEE) were used to account for potential correlation in mortality rates among patients sampled within hospital clusters (29). We examined an alternative to the inde-pendence assumption (i.e., no association with mortality rates) for within-hospital correlation using the quasi-likelihood independence criterion (30). An exchangeable correlation structure provided inferior fit to the independence working correlation structure, suggesting that a hospital effect was present in the model. All patient data in Tables 1 and 2 were considered to be potential candidate variables for the models. To test if hyperoxia exposure remained a significant indepen-dent predictor of in-hospital mortality when the propensity of individuals to be exposed to hyperoxia was adjusted for, a sensitivity analysis was performed using propensity scores and adding the result of the propensity score into the final multi-variable model (31). Finally, we tested for possible first-order interactions in those variables retained in the model. Statis-tical analyses were conducted using SPSS software version 20.0 (SPSS, Chicago, IL), and significance was judged when p value was less than 0.05. The current sample size had a power greater than 80% to detect a significant difference in propor-tions for our primary analysis, assuming our prespecified p value of less than 0.017. Our reporting of observational data conforms with Strengthening the Reporting of Observational Studies in Epidemiology guidelines (32).

RESUlTSDuring the 5-year period, there were 2,894 patients who met the inclusion criteria, which came from 84 different hospitals. All ICUs were nonneurologic/neurosurgical models. Diagno-sis included 554 AIS (19%), 936 SAH (32%), and 1,404 ICH (49%). There were more males in the cohort; patients were predominantly White and were from urban, large, or extralarge community (nonacademic) hospitals; and most of the patients were living independently before hospital admission. Do-not-resuscitate status at admission was present in less than 5% of the

cohort and not significantly different between exposed groups. Additional baseline characteristics of the cohort are shown in Table 1. In the cohort, 1,084 patients (38%) were normoxic, 1,316 (46%) were hypoxic, and 450 (16%) were hyperoxic. Of hypoxic patients, only two had Pao

2 less than 60 mm Hg (7.99

kPa) documented in ABG at admission to the ICU.There were no differences in terms of exposure to oxygen-

ation status in relation to neurologic diagnosis. The only sig-nificant comorbidity associated with oxygenation status was history of pulmonary disease, with more patients with this comorbidity being hypoxic. Hyperoxia exposure was more prevalent in community nonacademic and small to medium size hospitals. Hyperoxic patients tended to have lower GCS scores than hypoxic or normoxic patients (Table 1). Additional significant physiologic data over the first 24 hours of admis-sion for all exposed groups are shown in Table 2.

Overall 52% of patients met the primary outcome of in-hospital death at 28 days (Table 3). Mortality was higher in the hyperoxia group (268 of 450; 60% [95% CI, 57–65]) followed by the hypoxia group (690 of 1,316; 53% [95% CI, 49–55]) and the normoxia group (502 of 1,084; 47% [95% CI, 46–52]) (Table 3). The hyperoxia group had signif-icantly higher mortality compared with the normoxia group (OR, 1.7 [95% CI, 1.3–2.1]; p < 0.0001) and higher mor-tality compared with the hypoxia group (OR, 1.3 [95% CI 1.1–1.7]; p < 0.01). The hypoxia group had higher mortality when compared with the normoxia group (OR, 1.3 [95% CI, 1.1–1.5]; p < 0.01) (Table 3). On Kaplan-Meier analysis, the proportion of survivors at 28 days was significantly higher in the normoxia group as compared with the hyperoxia group (Fig. 2).

In the secondary analysis, hyperoxic SAH patients had higher mortality as compared with the normoxia group (OR, 2.5 [95% CI, 1.7–3.3]; p = 0.0001) and hypoxic (OR, 2.0 [95% CI, 1.3–2.5]; p < 0.002); and hyperoxic ICH patients had higher mortality as compared with the normoxia group (OR, 1.3 [95% CI, 1.03–1.7]; p = 0.02) (Table 3).

In the multivariable analysis, the following variables were found to be significantly associated with in-hospital mor-tality: age, GCS, ICH diagnosis, hypothermia at admission to ICU, hypotension or hypertension (MAP ≥ 120 mm Hg) organ dysfunctions, and exposure to hyperoxia. The following variables were significantly associated with survival: tempera-ture more than or equal to 37.5°C and ICP monitor place-ment. The multivariable analysis also suggested a significant effect related to a higher risk of death in hospitals with lower patient volume (Table 4; Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/CCM/A747). We did not observe significant interactions in the variables retained in the models.

Finally, we performed a sensitivity analysis to test whether hyperoxia remained a significant independent predictor of in-hospital mortality when adjusted for propensity scores (i.e., probability for hyperoxia exposure). We calculated each patient’s probability of hyperoxia exposure by fitting a logistic regression with hyperoxia as the dependent variable. We found that the

http://links.lww.com/CCM/A747

Rincon et al


following variables were significant independent predictors of hyperoxia exposure: male gender, non-ED boarder status, small or medium hospital size, hypothermia, hypotension or hyperten-sion (MAP < 60 mm Hg or MAP ≥ 120 mm Hg), tachycardia, bradypnea, higher hematocrit, and higher WBC count at admis-sion to the ICU (Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/CCM/A747). No other variables were significantly associated with hyperoxia exposure. The prob-abilities, expressed in quintiles, were included in the final GEE model (Table 4). In the absence of the calculated propensity score, the OR for in-hospital mortality related to hyperoxia was similar (OR, 1.3 [95% CI, 1.04–1.6]). These data indicate that hyperoxia remained a significant independent predictor of death when adjusted for the probability of being exposed to hyperoxia.

DISCUSSIONIn this large multicenter cohort study of ventilated stroke patients admitted to an ICU, we have demonstrated that expo-sure to hyperoxia was common and associated with lower likelihood of survival after hospital admission. Similarly, we demonstrated that after controlling for confounding variables and for hospital-specific effects in a robust multivariable anal-ysis, exposure to hyperoxia was an independent predictor of

in-hospital death. The effect of hyperoxia on in-hospital mor-tality was also independent after adjusting for the propensity or probability of exposure to hyperoxia. To our knowledge, this is the first study to report on the association between arte-rial hyperoxia and in-hospital mortality in ventilated stroke patients admitted to an ICU.

Traditionally, both arterial oxygenation (measured as Pao2) and

oxygen saturation levels by pulse oximetry are used as physiologic targets to guide ventilator management in the ICU (1). In lung-injured patients, common recommendations for oxygenation, based on results from large clinical trials, suggest a Pao

2 goal of 55–

80 mm Hg (7.33–10.7 kPa) (33). Current guidelines with regard to oxygen therapy for critically ill and neurologic patients suggest maintenance of normoxia with target saturation levels of 94–98% (1, 34). However, avoiding hypoxemia after brain injury to maintain adequate oxygen delivery to areas at risk is of crucial importance as well (1). To this end, supplementary oxygen is routinely adminis-tered to brain-injured patients, even in those with adequate oxygen saturation. Current day-to-day practices of oxygen administration in ventilated brain-injured patients may deviate from conventional recommendations for arterial Pao

2 or oxygen saturation targets

with the idea of increasing oxygen delivery or just too be on the “safe-side” (35).

TablE 2. Significant Physiologic and laboratory Characteristics of Ventilated Stroke Patients at admission to the ICU

Characteristics, Mean (sd) Total (n = 2,894) Hypoxia (n = 1,084) Normoxia (n = 1,316) Hyperoxia (n = 450) p

Temperature °C 36.3 (1.1) 36.3 (1.0) 36.2 (1.0) 36.2 (1.2) 0.001

Respiratory rate low per min 12 (4) 12 (4) 11 (3) 12 (4) 0.001

Respiratory rate high per min 22 (7) 22 (7) 21 (7) 22 (8) 0.0002

MAP lowest (mm Hg) 68 (19) 70 (18) 70 (18) 64 (21) < 0.0001

MAP highest (mm Hg) 116 (21) 116 (20) 117 (21) 118 (23) 0.04

Hematocrit lowest (%) 35 (7) 35 (6) 34 (6) 36 (7) < 0.0001

Hematocrit highest (%) 37 (6) 37 (6) 36 (6) 38 (7) < 0.0001

WBC lowest (per μL) 13 (13) 13 (7) 12 (5) 15 (25) 0.004

WBC highest (per μL) 15 (16) 14 (7) 14 (5) 17 (29) 0.001

Sodium(Na) highest mEq/L 140 (6) 141 (5) 141 (6) 142 (7) 0.02

Bicarbonate (HCO3) lowest mEq/L

22 (4) 23 (4) 22 (4) 22 (4) 0.0001

Bicarbonate (HCO3) highest mEq/L

23 (4) 24 (3) 24 (3) 24 (4) < 0.0001

Creatinine high (mg/dL) 1.3 (1.2) 1.3 (1.1) 1.2 (1.1) 1.4 (1.4) 0.04

Urine output, L/24 hr, median (interquartile range)

2.0 (1.2–3.3) 2.1 (1.3–3.3) 1.9 (1.2–3.1) 2.1 (1.3–3.5) 0.02

Pao2, mm Hg, mean (sd)/kPa (sd) 274 (148)/ 15.9 (7.2)

119 (54)/ 15.9 (7.2)

179 (48)/ 23.8 (6.4)

417 (79)/ 55.6 (10.5)

< 0.0001

Pao2/Fio2 289 (127) 203 (64) 403 (70) 459 (125) < 0.0001

Arterial pH (arterial blood gases) 7.43 (0.07) 7.43 (0.07) 7.43 (0.07) 7.42 (0.08) 0.02

MAP = mean arterial pressure.




Studies in healthy subjects have shown that hyperoxia is

associated with a decrease in cerebral blood flow by 11–33% (4,

5). It is likely that during abnormal conditions high oxygen con-

centrations are associated with decrease in delivery of oxygen

as well (36). On the other hand, observational studies in criti-cally ill populations have demonstrated the detrimental effects of hyperoxia (24) and that excessive oxygen may be deleterious for cardiac output, peripheral vascular resistances, cerebrovas-cular flow, and lung function (36, 37). The vasoconstricting effects of hyperoxia or hyperoxia-induced hypocapnia may not be widely recognized as potential adverse effects in brain-injured patients (9, 38). In this case, if perfusion decreases more than arterial oxygen content increases during hyperoxia, then regional oxygen delivery also decreases (38). This mechanism, and not just that attributed to reactive oxygen species (2, 3), is likely to contribute to the worse outcomes in critically ill patients routinely exposed to hyperoxia. In fact, this hypothesis may be supported by our results. We showed that hyperoxic patients with supranormal arterial Pao

2 levels had higher risk

of in-hospital mortality as compared with hypoxic patients, suggesting that the excess mortality may be on the basis of unnecessary hyperoxia exposure.

In our secondary analysis, we also demonstrated that this effect was significant in all comparisons except for the AIS population. Within the AIS population, the effect of hyperoxia was not significant when compared with normoxia. Potential explanations for the lack of significant results on our second-ary analysis for AIS may have been related to the smaller num-ber of AIS patients seen in our cohort or a true lack of effect of hyperoxia as compared with normoxia as evidenced by other large cohort studies (39). Experimental use of normobaric high oxygen doses has been carried in AIS populations (8, 17, 19).

TablE 3. Outcomes of Ventilated Stroke Patients by Exposure Groups: Primary and Subgroup analysis

Outcomes of Patients Total Hypoxia Normoxia Hyperoxiaa

Primary analysis

Overall, n 2,894 1316 1084 450

Dead, n (%; 95% CI) 1,479 (52; 50–54) 690 (53; 49–55) 502 (47; 45–52) 268 (60; 57–65)b

Secondary analysis

Ischemic stroke, n 554 268 202 75

Dead, n (%; 95% CI) 261 (48; 44–52) 123 (46; 34–48) 95 (47; 43–57) 43 (57; 47–63)c

Subarachnoid hemorrhage 936 401 383 135

Dead, n (%; 95% CI) 403 (44; 50–54) 176 (45; 41–47) 139 (38; 33–43) 80 (60; 52–68)d

Intracerebral hemorrhage 1,404 647 499 240

Dead, n (%; 95% CI) 812 (59; 56–62) 391 (61; 57–65) 268 (54; 50–58) 145 (61; 55–67)e

Overall hospital length of stay, median (interquartile range)

7 (2–14) 7 (2–14) 7 (3–14) 4 (1–12)f

aDefined as arterial Pao2 ≥ 300 mm Hg (39.99 kPa).bp < 0.0001 for comparison between normoxia and hyperoxia; p=0.01 for comparison between hypoxia and hyperoxia and between hypoxia and normoxia.cp = 0.13 for comparison between normoxia and hyperoxia; p=.08 for comparison between hypoxia and hyperoxia; and p=.8 for comparison between hypoxia and normoxiadp < 0.0001 for comparison between normoxia and hyperoxia; p = 0.002 for comparison between hypoxia and hyperoxia; and p = 0.04 for comparison between hypoxia and normoxiaep = 0.09 for comparison between normoxia and hyperoxia; p = 0.98 for comparison between hypoxia and hyperoxia; and p = 0.02 for comparison between hypoxia and normoxia.fp < 0.0001 Wilcoxon-Kruskal-Wallis test, all pairs significantly different (Wilcoxon method).

Figure 2. Kaplan-Meier analysis of 28-d mortality between hyperoxia (dashed line) and normoxia (continuous line) in ventilated stroke patients (log-rank < 0.0001).

Rincon et al


In a small study, Singhal et al (19) reported transient improve-ment in ischemic areas seen with MRI in patients with mild or moderate strokes receiving normobaric oxygen at 1.0 Fio

2

during the first 24 hours but without any meaningful impact on clinical outcome. In contradiction to the results of Singhal et al (19), a prior quasi-experimental study suggested that AIS patients may have been better off at 7 months when receiving room air rather than supplemental oxygen at 1.0 Fio

2 during

the first 24 hours after stroke (18). In addition, data from a small trial suggested that therapy with hyperbaric oxygen may have been harmful in AIS patients, and a systematic review found no evidence that hyperbaric oxygen improved outcomes after AIS or brain injury (8, 17). Although the results of studies of normobaric oxygen supplementation in AIS patients may be encouraging, at least in the short-term, this practice cannot be extrapolated to ventilated stroke populations. Based on our results and the preponderance of data suggesting a harmful

effect of hyperoxia in the AIS population (8, 17, 18), we still favor maintenance of normoxia until further studies of oxy-gen therapy within critically ill ventilated AIS populations help clarify this issue.

The results of our study may have potential implications for future research. Beyond avoiding hypoxia, no recommendation currently exists to maintain a certain arterial oxygen or satura-tion level in ventilated stroke patients beyond the guidelines for management of AIS (40) and expert opinions in the field of brain injury (1, 9). Common sense would call for avoidance of unnecessary arterial hyperoxia and a controlled reoxygen-ation strategy targeting an arterial oxygen saturation higher than 94% (1). However, our study finds that the prevalence of hyperoxia at admission to ICU in critically ill ventilated stroke patients was as high as one in five patients. Of these patients, more than half had Pao

2 more than or equal to 400 mm Hg. To

this end, hyperoxia appeared to be a common occurrence in

TablE 4. Multivariable analysis of Factors associated With In-Hospital Mortality in Ventilated Stroke Patients

Variable OR low High p

Intercepta 0.03 0.02 0.07 < 0.0001

Age (per yr) 1.02 1.02 1.03 < 0.0001

GCS (reference GCS > 12)

GCS < 8 7.56 5.40 10.57 < 0.0001

GCS 8–12 1.65 1.14 2.37 0.005

Diagnosis (reference ischemic stroke)

Subarachnoid hemorrhage 0.90 0.67 1.21 NS

Intracerebral hemorrhage 1.30 1.01 1.66 0.04

Admission temperature (reference normothermia 36–37.5°C)

Temperature ≥ 37.5°C 0.76 0.64 0.90 0.002

Hypothermia temperature < 36°C 1.52 1.26 1.83 < 0.0001

Admission MAP (reference normotension)

MAP ≥ 120 mm Hg 1.03 0.84 1.27 0.002

MAP < 60 mm Hg 1.62 1.29 2.03 < 0.0001

Intracranial pressure monitor placement 0.59 0.49 0.72 < 0.0001

Organ dysfunction (reference none)

Only one 1.31 1.07 1.61 0.008

Two or more 2.58 1.81 3.67 < 0.0001

Oxygen exposure (reference normoxia)

Hyperoxia (Pao2 ≥ 300 mm Hg [39.99 kPa]) 1.22 1.04 1.48 0.04

Hypoxia (Pao2 < 60 mm Hg [7.99 kPa] or Pao2/Fio2 < 300)

0.88 0.71 1.10 NS

Propensity score for hyperoxia (per quintile) 1.13 1.04 1.24 0.01

OR = odds ratio, GCS = Glasgow Coma Scale, NS = not significant, MAP = mean arterial pressure.aWe used generalized estimated equations under a robust estimation technique assuming an independence working correlation matrix. An exchangeable matrix provided inferior fit suggesting increased risk of death in hospitals with lower patient volume for all diagnoses. Hospitals with lower patient volume had higher mortality (Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/CCM/A747).




our study population. Our data provide insight into a potential prevalent problem applicable to the immediate resuscitation phase following stroke and provide important information to fill the current knowledge and evidence gap.

Our study has limitations. First, our analysis is observational in nature, limiting the inferences that can be made about causal relationships. Second, the inherent nature of the database used to perform our analysis was designed from an ICU perspective rather than to collect prospective data and endpoints commonly used in neurologic outcome research. We did not have data related to specific indexes of disease severity, surrogates of brain damage, or any other data related to specific treatments in stroke patients. This suggests the possibility of residual confounding, which may have influenced the association between hyperoxia and our outcome. In addition, the PI database did not capture ventilator data, which could have served as a surrogate of lung injury, and these could have also contributed to confound our association. Third, our study used a priori definitions of hyper-oxia as suggested by prior animal (26) and observational studies (24). Although our study suggests an association between hyper-oxia on in-hospital mortality, we were unable to define the pre-cise Pao

2 level at which the maximal risk for death would occur.

Similarly, our exposure groups were defined based on the first Pao

2 value measured within the first 24 hours at admission to the

ICU, and as ABG data within PI are not time stamped, we could not quantify the effect of duration of hyperoxia on mortality. Fourth, our data are limited by the quality of the PI repository, but the advantages of our robust sample size are potentially off-set by our inability to audit the data elements. Fifth, our cohort may have been subjected to selection bias in the sense that our sample may not be representative of the whole population under study or ICUs in the United States, so our results should be interpreted with this in mind. To this end, additional studies and clinical trials within specific populations of stroke patients may be needed to further expand our knowledge in the field and con-firm or refute our results. Finally, because the study focuses on nonneurologic or neurosurgical ICUs in the United States, our results may not be generalizable to other settings where man-agement of ventilated stroke patients and patient characteristics may differ substantively.

CONClUSIONSIn ventilated stroke patients admitted to the ICU, arterial hyperoxia was independently associated with in-hospital death as compared with either normoxia or hypoxia. These data underscore the need for studies of controlled reoxygenation in ventilated critically ill stroke populations. In the absence of results from clinical trials, unnecessary oxygen delivery should be avoided in ventilated stroke patients.

aCKNOWlEDGMENTSWe thank Andrew Kramer, PhD, and the Cerner Corporation for permitting access to the Project IMPACT database and Paul Lanken, MD, Division of Pulmonary, Allergy & Critical Care,

University of Pennsylvania Medical Center, for the insightful comments and review of the manuscript.

REFERENCES 1. Seder DB, Riker RR, Jagoda A, et al: Emergency neurological life

support: Airway, ventilation, and sedation. Neurocrit Care 2012; 17(Suppl 1):S4–20

2. Hess ML, Manson NH: Molecular oxygen: Friend and foe. The role of the oxygen free radical system in the calcium paradox, the oxygen paradox and ischemia/reperfusion injury. J Mol Cell Cardiol 1984; 16:969–985

3. Rousseau A, Tesselaar E, Henricson J, et al: Prostaglandins and radi-cal oxygen species are involved in microvascular effects of hyperoxia. J Vasc Res 2010; 47:441–450

4. Floyd TF, Clark JM, Gelfand R, et al: Independent cerebral vasocon-strictive effects of hyperoxia and accompanying arterial hypocapnia at 1 ATA. J Appl Physiol 2003; 95:2453–2461

5. Johnston AJ, Steiner LA, Gupta AK, et al: Cerebral oxygen vasore-activity and cerebral tissue oxygen reactivity. Br J Anaesth 2003; 90:774–786

6. Farquhar H, Weatherall M, Wijesinghe M, et al: Systematic review of studies of the effect of hyperoxia on coronary blood flow. Am Heart J 2009; 158:371–377

7. Rawles JM, Kenmure AC: Controlled trial of oxygen in uncomplicated myocardial infarction. Br Med J 1976; 1:1121–1123

8. Rusyniak DE, Kirk MA, May JD, et al; Hyperbaric Oxygen in Acute Ischemic Stroke Trial Pilot Study: Hyperbaric oxygen therapy in acute ischemic stroke: Results of the hyperbaric oxygen in acute ischemic stroke trial pilot study. Stroke 2003; 34:571–574

9. Diringer MN: Hyperoxia: Good or bad for the injured brain? Curr Opin Crit Care 2008; 14:167–171

10. Yusa T, Beckman JS, Crapo JD, et al: Hyperoxia increases H2O2 pro-duction by brain in vivo. J Appl Physiol 1987; 63:353–358

11. Mickel HS, Vaishnav YN, Kempski O, et al: Breathing 100% oxygen after global brain ischemia in Mongolian Gerbils results in increased lipid peroxidation and increased mortality. Stroke 1987; 18:426–430

12. Haelewyn B, Chazalviel L, Nicole O, et al: Moderately delayed post-insult treatment with normobaric hyperoxia reduces excitotoxin-induced neuronal degeneration but increases ischemia-induced brain damage. Med Gas Res 2011; 1:2

13. Singhal AB, Dijkhuizen RM, Rosen BR, et al: Normobaric hyperoxia reduces MRI diffusion abnormalities and infarct size in experimental stroke. Neurology 2002; 58:945–952

14. Henninger N, Bouley J, Nelligan JM, et al: Normobaric hyperoxia delays perfusion/diffusion mismatch evolution, reduces infarct vol-ume, and differentially affects neuronal cell death pathways after suture middle cerebral artery occlusion in rats. J Cereb Blood Flow Metab 2007; 27:1632–1642

15. Shin HK, Dunn AK, Jones PB, et al: Normobaric hyperoxia improves cerebral blood flow and oxygenation, and inhibits peri-infarct depolar-izations in experimental focal ischaemia. Brain 2007; 130:1631–1642

16. Fujiwara N, Mandeville ET, Geng X, et al: Effect of normobaric oxy-gen therapy in a rat model of intracerebral hemorrhage. Stroke 2011; 42:1469–1472

17. Alternative Therapy Evaluation Committee for the Insurance Corporation of British Columbia: A review of the scientific evidence on the treatment of traumatic brain injuries and strokes with hyperbaric oxygen. Brain Inj 2003; 17:225–236

18. Rønning OM, Guldvog B: Should stroke victims routinely receive sup-plemental oxygen? A quasi-randomized controlled trial. Stroke 1999; 30:2033–2037

19. Singhal AB, Benner T, Roccatagliata L, et al: A pilot study of nor-mobaric oxygen therapy in acute ischemic stroke. Stroke 2005; 36:797–802

20. Rincon F, Mayer SA, Rivolta J, et al: Impact of delayed transfer of criti-cally ill stroke patients from the emergency department to the neuro-ICU. Neurocrit Care 2010; 13:75–81

Rincon et al


21. Le Gall JR, Lemeshow S, Saulnier F: A new Simplified Acute Physiology Score (SAPS II) based on a European/North American multicenter study. JAMA 1993; 270:2957–2963

22. Trzeciak S, Jones AE, Kilgannon JH, et al: Significance of arterial hypotension after resuscitation from cardiac arrest. Crit Care Med 2009; 37:2895–2904

23. Diringer MN, Edwards DF: Admission to a neurologic/neurosurgical intensive care unit is associated with reduced mortality rate after intra-cerebral hemorrhage. Crit Care Med 2001; 29:635–640

24. Kilgannon JH, Jones AE, Shapiro NI, et al; Emergency Medicine Shock Research Network (EMShockNet) Investigators: Association between arterial hyperoxia following resuscitation from cardiac arrest and in-hospital mortality. JAMA 2010; 303:2165–2171

25. Halpern NA, Pastores SM, Thaler HT, et al: Changes in critical care beds and occupancy in the United States 1985-2000: Differences attributable to hospital size. Crit Care Med 2006; 34:2105–2112

26. Douzinas EE, Patsouris E, Kypriades EM, et al: Hypoxaemic reperfu-sion ameliorates the histopathological changes in the pig brain after a severe global cerebral ischaemic insult. Intensive Care Med 2001; 27:905–910

27. Dellinger RP, Levy MM, Rhodes A, et al; Surviving Sepsis Campaign Guidelines Committee Including the Pediatric Subgroup: Surviving sepsis campaign: International guidelines for management of severe sepsis and septic shock: 2012. Crit Care Med 2013; 41:580–637

28. Ranieri VM, Rubenfeld GD, Thompson BT, et al; ARDS Definition Task Force: Acute respiratory distress syndrome: The Berlin defini-tion. JAMA 2012; 307:2526–2533

29. Localio AR, Berlin JA, Ten Have TR, et al: Adjustments for center in mul-ticenter studies: An overview. Ann Intern Med 2001; 135:112–123

30. Ziegler A, Vens M: Generalized estimating equations. Notes on the choice of the working correlation matrix. Methods Inf Med 2010; 49:421–426

31. Austin PC: An introduction to propensity score methods for reduc-ing the effects of confounding in observational studies. Multivariate Behav Res 2011; 46:399–424

32. von Elm E, Altman DG, Egger M, et al; STROBE Initiative: The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: Guidelines for reporting observational studies. Lancet 2007; 370:1453–1457

33. Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syn-drome. The Acute Respiratory Distress Syndrome Network. N Engl J Med 2000; 342:1301–1308

34. O’Driscoll BR, Howard LS, Davison AG: BTS guideline for emer-gency oxygen use in adult patients. Thorax 2008; 63(Suppl 6):vi1–68

35. de Jonge E, Peelen L, Keijzers PJ, et al: Association between adminis-tered oxygen, arterial partial oxygen pressure and mortality in mechan-ically ventilated intensive care unit patients. Crit Care 2008; 12:R156

36. Lodato RF: Decreased O2 consumption and cardiac output dur-ing normobaric hyperoxia in conscious dogs. J Appl Physiol 1989; 67:1551–1559

37. Zwemer CF, Whitesall SE, D’Alecy LG: Hypoxic cardiopulmonary-cerebral resuscitation fails to improve neurological outcome following cardiac arrest in dogs. Resuscitation 1995; 29:225–236

38. Iscoe S, Beasley R, Fisher JA: Supplementary oxygen for nonhypox-emic patients: O2 much of a good thing? Crit Care 2011; 15:305

39. Young P, Beasley R, Bailey M, et al; Study of Oxygen in Critical Care (SOCC) Group: The association between early arterial oxygenation and mortality in ventilated patients with acute ischaemic stroke. Crit Care Resusc 2012; 14:14–19

40. Adams HP Jr, del Zoppo G, Alberts MJ, et al: Guidelines for the early man-agement of adults with ischemic stroke: A guideline from the American Heart Association/American Stroke Association Stroke Council, Clinical Cardiology Council, Cardiovascular Radiology and Intervention Council, and the Atherosclerotic Peripheral Vascular Disease and Quality of Care Outcomes in Research Interdisciplinary Working Groups: The American Academy of Neurology affirms the value of this guideline as an educa-tional tool for neurologists. Stroke 2007; 38:1655–1711

association between hyperoxia and mortality after stroke- a multicenter cohort study.pdf

Documents