Neurological Research A Journal of Progress in Neurosurgery, Neurology and Neurosciences
ISSN: 0161-6412 (Print) 1743-1328 (Online) Journal homepage: http://www.tandfonline.com/loi/yner20
The effect of normobaric oxygen in patients with acute stroke: a systematic review and metaanalysis Jiayue Ding, Da Zhou, Meng Sui, Ran Meng, Ankush Chandra, Jie Han, Yuchuan Ding & Xunming Ji To cite this article: Jiayue Ding, Da Zhou, Meng Sui, Ran Meng, Ankush Chandra, Jie Han, Yuchuan Ding & Xunming Ji (2018): The effect of normobaric oxygen in patients with acute stroke: a systematic review and meta-analysis, Neurological Research, DOI: 10.1080/01616412.2018.1454091 To link to this article: https://doi.org/10.1080/01616412.2018.1454091
View supplementary material
Published online: 30 Mar 2018.
Submit your article to this journal
Article views: 13
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=yner20
Neurological Research, 2018 https://doi.org/10.1080/01616412.2018.1454091
REVIEW
The effect of normobaric oxygen in patients with acute stroke: a systematic review and meta-analysis Jiayue Dinga,b# , Da Zhoua,b# , Meng Suic, Ran Menga,b , Ankush Chandrad,e , Jie Hanf, Yuchuan Dingd and Xunming Jib,g a
Department of Neurology, Xuanwu Hospital, Capital Medical University, Beijing, China; bBeijing Institute for Brain Disorders, Beijing, China; Department of Economics, Fordham University, Bronx, NY, USA; dDepartment of Neurosurgery, Wayne State University School of Medicine, Detroit, MI, USA; eDepartment of Neurosurgery, University of California San Francisco, San Francisco, CA, USA; fDepartment of Neurology, The First Affiliated Hospital of Dalian Medical University, Dalian, China; gDepartment of Neurosurgery, Xuanwu Hospital, Capital Medical University, Beijing, China c
ABSTRACT
Background: Normobaric oxygen (NBO) has received considerable attention due to controversial data in brain protection in patients with acute stroke. This study aims to analyze current data of NBO on brain protection as used in the clinic. Methods: We searched for and reviewed relevant articles and references from Pubmed, Medline, Embase, Cochrane, and Clincialtrials.gov that were published prior to October 2017. Data from prospective studies were processed using RevMan5.0 software, provided by Cochrane collaboration and transformed using relevant formulas. Results: A total of 11 prospective RCT studies including 6366 patients with acute stroke (NBO group, 3207; control group, 3159) were enrolled in this analysis. △NIHSS represented the values of NIHSS at 4, 24 h, or 7 days post-stroke minus baseline NIHSS. Compared to controls, there was a minor trend toward NBO benefits in short-term prognostic indices, as indicated by decreased ΔNIHSS at our defined time points. By contrast, NBO decreased Barthel Index scores between 3 and 7 months, and increased death rates at 3, 6 months, and 1 year, whereas, modified Rankin Scale scores between 3 and 6 months were unchanged. Conclusions: The existing trends toward benefits revealed in this meta-analysis help us appreciate the promising value of NBO, although current evidence of NBO on improving clinical outcomes of stroke is insufficient. Well-designed multi-center clinical trials are encouraged and urgently needed to further explore the efficacy of NBO on brain protection. Abbreviations: BBB:Blood–brain barrier; NBO: Normobaric oxygen; HBO: Hyperbaric oxygen; PtO2: Tissue oxygen tension; RCTs: Randomized controlled trials; FiO2: Fraction of inspiration O2; BI scores: Barthel Index scores; mRS scores: Modified Rankin Scale scores; NIHSS: National Institute of Health Stroke Scale; SSS: Scandinavian stroke scores; DWI: Diffusion weighted imaging; ADC: Apparent diffusion coefficient; MTT: Mean transit time; RR: Relative risk; MD: Mean difference; NE: Experimental sample size; NC: Controlled sample size; SD: Standard deviation; M: Median; IQR: Interquartile range; ROS: Reactive oxygen species; eNOS: Endothelial nitric oxide synthase; NO: Nitric oxide; CBV: Cerebral blood volume; iNBO: Intermittent normobaric hyperoxia; pNBO: persistent normobaric hyperoxia; DSMB: Data and Safety Monitoring Board
Introduction Oxygen supplementation is a common adjuvant therapy for various diseases and has recently received considerable attention due to its controversial effects on brain protection in humans with acute stroke. Several animal studies have demonstrated the benefit effects of oxygen therapy in improving tissue oxygenation, promoting aerobic metabolism, ameliorating blood– brain barrier (BBB) disruption, and reducing peri-infarct depolarization, thereby, preventing the brain from ischemia-reperfusion injury [1,2]. Clinical trials have shown similar results corroborating the data
KEYWORDS
Oxygen inhalation therapy; stroke; meta-analysis; normobaric oxygen
from scientific studies [3–9]. On the contrary, other experimental [10,11] and clinical studies [12–15] have revealed that oxygen enhances lipid peroxidation and oxidative stress response, resulting in high rates of death and disability in animal models and patients. In addition, passive oxygen inhalation may increase the risks of pulmonary atelectasis, pulmonary edema and pneumonia [16]. Based on the oxygen pressure inhaled (equal to or more than atmosphere pressure, 1 ATM = 101.325 kPa), oxygen intervention can be categorized into two types: normobaric oxygen (NBO) and hyperbaric oxygen (HBO) [17].
CONTACT Ran Meng
[email protected] # These contributed equally to this article. The supplemental data for this article can be accessed at https://doi.org/10.1080/01616412.2018.1454091. © 2018 Informa UK Limited, trading as Taylor & Francis Group
ARTICLE HISTORY
Received 8 December 2017 Accepted 7 March 2018
2
J. DING ET AL.
HBO can increase brain tissue oxygen tension (PtO2) and confer protection against severe ischemia [17,18]. However, implementation requirements and side effects on the lung largely limit its use [10,11]. Meanwhile, NBO has gained significant attention recently and has become a research hotspot in the field of stroke, as it is safe and convenient, and can be initiated within a few minutes after stroke [19,20]. Notably, the efficacy of NBO on brain protection in stroke is still under debate [21,22]. Some researchers suggest that NBO is more adequate to be used in stroke patients during acute phase; while, certain clinical trials imply NBO not to be a valid strategy [3–9,12,13,15,23]. Our study aims to perform a secondary analysis of the current available data on NBO and its role in brain protection, especially, in a clinical scenario. Findings from our study will facilitate the understanding of the protective effects of NBO in acute stroke and may also provide clinical practitioners a guideline in treating stroke patients with NBO.
Methods Systematic Reviews and Meta-Analyses were performed in this study according to the Cochrane systematic review and the Preferred Reporting Items (Details of checklists are shown in supplement material 1) [24].
performed continuously in the acute phase (less than 48 h from stroke onset) for 21% at 1 ATM [20],
NEUROLOGICAL RESEARCH
Figure 1. Flowchart of included studies.
3
Figure 2. Summary of risk of bias.
meta-analysis. Otherwise we performed meta-analysis through random-effects model (I2 > 50%). Publication bias was shown using a funnel plot. It was recommended to use subgroup analysis to make the heterogeneity diminished. For each of the cases (calculation process shown in supplement material 2): (1) Given P, MD, experimental (NE), and control (NC) sample size from the raw study, we obtained standard deviation (SD) using a transformation function [26]: input formula ‘ = tinv (P, NE + NC-2)’ into Excel to arrive t value (NE + NC-2: degree of freedom); SE = MD/t (SE, standard error of MD); SD = SE×(1/ NE + 1/NC)−1/2. (2) When only median (M), interquartile range (IQR), and the original data were given, as long as the sample size was large enough (n > 25) [27]: mean = M and SD = IQR/1.35. (3) Given M and Range (a–b, R) only, we computed the ‘m’ and ‘SD’ using the following equation (m as mean) [28]: (a) If n ≤ 25, m = (a + 2M + b)/4; if n > 25, m = M; (b) If n ≤ 15, SD = {1/12[(a + 2 M + b)2/4+(b–a)2]}1/2; if 15 70, SD = R/6.
Results Search results After removing duplicates, irrelevant articles, and publications which that did not fulfill our PICOS criteria, a total of 11 RCTs published before October 2017, including 6366 patients (3207 in NBO group and 3159 in control) who came from seven independent patient groups and fulfilled the inclusion and analytical criteria of the study, were included in this analysis [3–9,12,13,15,23] (Figure 1). The currently largest RCT provided by Roffe et al. recruited almost 8000 patients and divided these subjects into three groups: continuous oxygen, nocturnal
Figure 3. Publication bias from the involved studies.
oxygen, and control [15]. We only channeled the sample size in continuous oxygen and control groups into our study. Moreover, the Oxford Center for Evidence-Based Medicine scale in each study was 2b or more than 2b. This scale was used to guarantee high-quality articles included in our meta-analysis. Characteristics of the 11 RCTs In the 11 RCTs mentioned above, studies conducted by Roffe et al. (2011) and Ali et al. (2014) were identified from the same sample [5,7]; Singhal et al. (2005 and 2007), González et al. (2010), and Wu et al. (2012) studies were from the same cohort of patients [3,6,8,9]. According to the Cochrane Collaboration’s tool for assessing the risk of bias, we found seven studies with low risk of bias (Roffe 2017, Ali 2014, Wu 2012, Roffe 2011, González 2010, Singhal 2005, and Singhal 2007) [3,5–9,15], two with unclear risk of bias (Mazdeh 2015 and Padma 2010) [4,12], and two with high risk of bias (Singhal 2010 and Rønning 1999) [13,23]. Selective bias is shown in Figure 2. Publication bias was processed using a funnel plot (Figure 3). Symmetrical graph indicated that the publication bias of these involved studies was low. According to the outcomes, we evaluated the effect
4
J. DING ET AL.
Table 1. Characteristics of included 11 RCTs. Study Roffe, 2017 [15]
Design RCT
Population# 2668/2668
Mazdeh, 2015 [4]
RCT
26/25
Ali, 2014 [5]**
RCT
148/141
Wu, 2012 [6]*
RCT
10/6
Roffe, 2011 [7]
RCT
148/141
Singhal, 2010 [23]
RCT
43/42
González, 2010 [9]*
RCT
8/6
Padma, 2010 [12]
RCT
20/18
Singhal, 2007 [3]*
RCT
4/2
Singhal, 2005 [8]
RCT
9/7
Rønning, 1999 [13]
RCT
292/258
Major outcome indicator mRS scores; BI scores; Mortality rate; Oxygen saturation; NIHSS etc. BI scores; mRS scores; Mortality rate etc. mRS scores; BI scores; Mortality rate etc. Lesion volume.; DWI/ADC/MTT; CBF/CBV etc. Oxygen saturation; NIHSS; GCS etc. NIHSS; Lesion volumes etc. DWI; MTT; rADC etc. NIHSS; mRS scores; BI scores; DWI etc. DWI/ADC; lactate; N-acetyl-aspartate etc. NISS; SSS; mRS scores; Lesion volume; DWI/ADC/MTT; CBF/CBV Penumbral salvage etc. SSS; BI scores; Mortality rate etc.
No improved
Efficacy
Bias evaluation Low
Beneficial
Unclear
Beneficial but not significantly
Low
Beneficial
Low
Beneficial
Low
No improved
High
Beneficial
Low
No improved
Unclear
Beneficial
Low
Beneficial
Low
Harmful
High
*
Participants and study details are from ‘Singhal 2005’.; Participants and study details are from ‘Roffe 2011’.; # Data are given as Treatment/Control. **
of NBO as ‘beneficial’, ‘no improvement’, and ‘harmful’. The characteristics of the RCTs are listed in Table 1. The modes of NBO intervention in the 11 RCT studies The modes of NBO intervention in the 11 studies were not the same, and this may influence the efficacy of NBO. Details about the FiO2 or flow velocity varied across studies. For example, FiO2 ≤ 50% was used in
three studies, while FiO2 > 50% was used in another two studies. Moreover, some provided oxygen flow velocities only instead of FiO2. The oxygen flow velocity applied in different studies was also inconsistent: the oxygen flow velocity was set as 2–3L/min in four studies, 4–10L/ min in two other studies, and 10L/min in the remaining five studies. To maintain consistency, data of FiO2 and flow velocity were converted according to the formula (FiO2 = 21 + 4 × Flow velocity), in attempt to facilitate the comparison.
Table 2. The modes of NBO intervention in the 11 RCTs. Study Treatment/Control Roffe, 2017 [15] Mazdeh, 2015 [4] Roffe, 2011 [7]; Ali, 2014 [5] Singhal, 2005 [8]; Singhal, 2007 [3]; González, 2010 [9]; Wu, 2012 [6] Singhal, 2010 [23] Padma, 2010 [12] Rønning, 1999 [13]
FiO2
Flow velocity (L/min)
Device
29–33%*/RA
2–3/RA#
Nasal cannula/RA
50%/RA 29–33%*/RA NA/RA
7.25*/RA 2–3/RA# 45/RA#
Venturi mask/RA Nasal cannula/RA Simple facemask/RA
NA/RA 61%*/RA 100%/RA
30–45/30–45 10/RA# 3/RA
Facemask/facemask Simple face mask/RA Nasal cannula/RA
Begin time Within 24 h after stroke onset mostly Within 12 h after stroke onset Within 24 h after stroke onset Within 12 h after witnessed symptom onset or 15 h after last seen neurologically intact Within 9 h after stroke onset Within 12 h after stroke onset Within 24 h after stroke onset
Duration 72 h/RA 12 h/RA 72 h/RA 8 h/RA 8 h/8 h 12 h/RA 24 h/RA
Notes: RA, room air; NA, data not applicable. * This data are not provided originally, but we get the result via calculation (the formula: FiO2 = 21 + 4 × flow velocity).; #oxygen only when clinically indicates.
– – 46(15.8)/48(18.6) – – 42(26)/43(25)
Patient Demographics in the selected 11 RCTs Patient age in the 11 RCTs ranged from 26 to 97 years, with the mean age of 72.3 ± 12.7 years. One article only presented the range of age, thus, we used the formula: [(the largest value + the smallest value)/2] to estimate the mean age. One article did not present the age of patients in the control group, and thus, we used the sample size of patients in the NBO group to estimate the whole average age. Data in NBO groups were well matched with controls in the 11 RCTs regarding the initial time of intervention recorded, the type of stroke (according to the TOAST criteria), and the severity of baseline neurological defects (such as NIHSS scores). Details are displayed in Table 3. Outcomes of patients in the 11 RCTs
Notes: SD, standard deviation; #Data are estimated value due to absent original information; –, information not produced.
– 58.2/53.9 – – – 37(12.7)/29(12.7) – 8.5(2.2) 7.6/9.8# 16(37.2)/25(59.5) – 166(56.8)/126(48.9) 74.1(14)/73.4(14) 55.8(13.2) 76.7(7)/76.1(8)
5
Furthermore, the devices used for NBO intervention varied significantly, including nasal cannula, simple facemask, and Venturi mask. Besides, patients directly inhaled the room air in the control group and did not wear facemasks as in a study by Singhal et al. (2010) (Table 2).
12.2(7)/12.1(6) 14.25/12.7 –
638(24)/684(26) – 34(23)/19(14) 6(66.7)/5(71.4) – – 39(15)/42(18) 27/32 – 42.5(15.8)/42(14.9) – – 1466(55)/1466(55) 14(53.8)/14(56) 65(43.9)/72(51.1) 4(44.4)/3(42.9)
Treatment/Control Roffe, 2017 [15] Mazdeh, 2015 [4] Roffe, 2011 [7]; Ali, 2014 [5] Singhal, 2005 [8]; Singhal, 2007 [3]; González, 2010 [9]; Wu, 2012 [6] Singhal, 2010 [23] Padma, 2010 [12]# Rønning, 1999 [13]
72.0(13)/72.0 (13) 40–70 73.0(12.0)/71.0(12.0) 67.0(16)/70.0(18)
20.7(10.1)/20.8(10.1) – 18(9)/16.5(8) –
185(6.9)/196(7.3) 8(30.8)/5(20) 9(6)/15(11) –
5(6)/5(6) – 6(7)/5(7) 14/11
Atrial fibrillation; n (%) Barthel index scores; mean (SD) Age; mean (SD) Study
Table 3. Baseline characteristics of patients in the 11 RCTs.
Male gender; n (%)
Begin time; mean (SD)
Hemorrhagic stroke; n (%)
NIHSS; median (IQR)
SSS; median (IQR)
NEUROLOGICAL RESEARCH
There was significant heterogeneity in patient outcomes in the 11 RCTs. In one study, BI was higher in the NBO group than in controls, while in four studies, the results were opposite. Rønning et al. found that seven-month BI tended to be lower than that of controls in severe patients after NBO treatment, whereas, this tendency did not appear in mild or moderate patients. In two studies, the mRS scores in the NBO group were almost the same as that in the control, especially in the largest RCTs performed by Roffe et al. in 2017. However, in another three studies, the mRS scores were obviously lower in the NBO group than controls. Two prospective studies recruiting a small number of subjects were completed by Singhal et al. in 2005 and 2010, respectively. The findings in 2005 showed a benefit tendency from NBO, as measured by 4 and 24-h NIHSS scores after stroke (24-h NIHSS scores with statistical significance). Nonetheless, no benefit was obtained from the study in 2010. Roffe et al. in 2011 and Singhal et al. in 2005 demonstrated that NBO improved NIHSS scores at 1 week. Additionally, the data from Singhal et al. 2005 revealed that NIHSS and SSS tended to improve at 3 months among NBO-treated patients after stroke. On the contrary, Padma et al. found the opposite results. The multi-center trial in 2017, displayed no difference in seven-day NIHSS scores between NBO and control groups. The ratio of death in patients with NBO was almost equal to that in controls in two studies [4,7,13], while it was higher than controls in another three studies [5,23]. Rønning et al. demonstrated that NBO might
3-mo, 73.05/73.80 7mo, 95(22.2)/100(14.8)#
Padma, 2010 [12] Rønning, 1999 [13]
3-mo, 2/2.2 –
–
3-mo, 2.53(1.83)/2.52(3.4) 6-mo, 2.73(2.27)/3.28(2.01)* 6-mo, 2.96(2.22)/2.93(2.96) 3-mo, 3.2(2.2)/4.1(1.6)
mRS; mean (SD)
– 1 year, 91(31.2)/70(27.1)
3-mo, 1(2.3)/1(2.4)
3-mo, 222(9.2)/214(8.8) 6-mo, 5(19.2)/3(12) 6-mo, 22(14.9)/21(14.9) –
Mortality rate; n(%)
4 h, −0.37(2.78)/–0.43(3.32)# 24 h, 0.17(4.15)/–0.73(5.24) 1-wk, −2.65(0.64)/–3.2(0.90); –
1-wk, −5(4.11)/1.75 (4.78)*
24 h, −5.5(3.67)/4(5.08)*
1-wk, −5(0.14)/–5(0.14)# – 1-wk, −2(2.22)/–1(1.48)* 4 h, −3.25(3.9)/2.75(4.9)*,#
ΔNIHSS; mean (SD)
– 7 mo, 54/55
–
– – – 3-mo, 47/32
SSS; median
– – – 4 h and 24 h DWI, 11(0.5)/ 9.3(1.2)*, 8.9(0.3)/ 2.3(0.8) ADC, 0.10(0.3)/–4.3(1.3), 1.0(0.2)/–9.1(1.0) CBV, −0.02(0.2)/0.7(0.1), −0.2(0.08)/0.5(0.1) CBF, −0.4(0.2)/–0.9(0.2), 0.2(0.09)/– 0.7(0.1) MTT, 0.9(0.2)/0.5(0.2), 1.3(0.1)/0.5(0.1) NAA no change, Lac↓/NAA↓, Lac↑ 4 h and 24 h Percent lesion growth(%): 100(78)/82(80), 20(31)/16(28) – –
Imaging manifestations; mean (SD)
Note: Barthel index scores (10-item, 100-point) = [Barthel index scores(3-itemt) × 2.39] + 0.14. * The difference between treatment and control is statistically significant.; #Data are estimated value which are converted by relevant formula from original information, details shown in supplementary material 2.
–
3-mo, 70.2(28.1)/70.9(28.1)# 6-mo, 64.6(35.2)/56.9(32.8) 6-mo, 71.6(26.6)/77.3(17.7)# –
Barthel index scores; mean (SD)
Singhal, 2010 [23]
Treatment/Control Roffe, 2017 [15] Mazdeh, 2015 [4] Roffe, 2011 [7]; Ali, 2014 [5] Singhal, 2005 [8]; Singhal, 2007 [3]; González, 2010 [9]; Wu, 2012 [6]
Study
Table 4. Outcomes of patients in the 11 RCTs.
6 J. DING ET AL.
NEUROLOGICAL RESEARCH
7
Figure 4. Meta-analysis of the differences of ΔNIHSS between the NBO group and the control group at 4, 24 h, and 7 days after stroke.
Figure 5. Meta-analysis of the differences of BI between the NBO group and the control group at 3, 6 and 7 months after stroke.
not be beneficial and might even be harmful to patients with severe stroke. For instance, the one-year survival rates for the NBO group and the control group were 81.8 and 90.7% in mild to moderate stroke cohort, respectively (p = 0.02), and 53.4 and 47.7% in severe stroke cohort, respectively (p = 0.60). Data are shown in Table 4. Pooled analyses Meta-analysis was performed with RevMan5.0 software provided by Cochrane Collaboration website. There was no heterogeneity among studies when analyzing
BI, mRS, and mortality (χ2 = 5.29, p = 0.15, I2 = 43%; χ2 = 1.76, p = 0.62, I2 = 0%; χ2 = 0.91, p = 0.92, I2 = 0%), indicating that the fixed-effects model can be used to determine pooled effect size (I2 ≤ 50%). In contrast, significant heterogeneity existed when analyzing the ΔNIHSS at three time points (NIHSS at 4, 24 h, and 7 days after stroke minus baseline NIHSS) in five studies (χ2 = 6.60, p
