Study selection and characteristics
The PRISMA statement flowchart (Fig. 1) shows the detailed process of literature screening, study selection, and reasons for exclusion. Ultimately, 1197 TBI patients (611 treated with EPO, 586 treated with placebo) from seven RCTs were included in this meta-analysis (Table 1).
Table 1
Characteristics of randomized controlled trials included in the meta-analysis
Study/Year | Country | Study design | No. of patients | Mean age, y | Male patients, % | EPO vs. Control | Clinical Setting | Interventions | Time from trauma to intervention | Control | Follow-up Duration | Outcome Measures |
Nirula et al, 2010 | USA | Single-center, double-blind RCT | 16 | 36.5 | 68.6 | 11 vs. 5 | Severe or Moderate TBI: Blunt trauma patients with evidence of TBI on CT, GCS < 13 | EPO 40,000 IU IV | Within 6 hours | Equal volume 0.9% NaCl | Discharge from Hospital or dead | Serum s-100B, NSE levels, ICP values, adverse events |
Abrishamkar et al, 2012 | Iran | double-blind RCT | 54 | 26.3 | 100 | 27 vs. 27 | Severe TBI w/DAI: GCS 4–8, exclude hemorrhage | rhEPO 2,000 IU SC for six doses in two weeks (on days: 2, 4, 6, 8 and 10) | Average time was 5 hours | Equal volume 0.9% NaCl | Discharge from Hospital or dead | Mean GOS score, length of hospital stay, mortality |
Robertson et al, 2014 | USA | Multi-center, double-blind RCT | 200 | 29.5 | 86.5 | 102 vs. 98 | TBI: Closed head injury who were unable to follow commands | Regimen 1: EPO 500 IU/kg per dose IV,1 dose given within 6 hours of injury followed by 2 additional doses given every 24 hours; regimen 2: EPO 1 dose given within 6 hours of injury | Within 6 hours | Equal volume 0.9% NaCl over 2 minutes for each dose | 6 months | GOS score, DRS Score, adverse events |
Aloizos et al, 2015 | Greece | Single-center, open-label RCT | 42 | 36.7 | 92.9 | 24 vs. 18 | Severe TBI: Blunt head trauma, GCS < 9, or hypotension (systolic blood pressure < 100 mmHg) | EPO 10,000 IU SC for 7 consecutive days | Within 6 hours | No placebo treatment | 6 months | GOS-E Score, length of ICU stay, adverse events |
Nichol et al, 2015 | Australia, New Zealand, France, Germany, Finland, Ireland, and Saudi Arabia | Multi-center, double-blind RCT | 606 | 30.5 | 83.3 | 305 vs. 298 | Severe or moderate TBI: non-penetrating traumatic brain injury (GCS 3–12) | Epoetin alfa 40,000 IU SC once per week for a maximum of three doses | Within 24 hours | Equal volume 0.9% NaCl, once per week for a maximum of three doses | 6 months | GOS-E Score, adverse events |
Li et al, 2015 | China | Single-center, open-label RCT | 159 | 42.3 | 61.6 | 78 vs. 80 | Severe TBI: head trauma plus the worst initial GCS of 7 or less on arrival. | rhEPO subcutaneous injection on the admission day (within 2 h of admission), and on day 3,6,9 and 12 after admission, daily dose of 100 units/kg (average 5999 units) | within 5 hours | received the same volumes of subcutaneous normal saline on the admission day and on day 3,6,9 and 11 after admission | 2 months | GOS score, Serum s-99B, NSE levels, Blood pressure, hemoglobin levels, adverse events |
Bai et al, 2018 | China | Single-center, open-label RCT | 120 | 43.8 | 70.8 | 60 vs. 60 | Severe TBI: GCS score less than 8 | rhEPO 6000 IU subcutaneous injection on the admission day (within 2 h of admission), and on day 3,5,10 and 15 after admission | NaN | received the same volumes of subcutaneous normal 0.9% saline on the same time | 10 weeks | GOS score, mortality, adverse events |
EPO, erythropoietin; TBI, traumatic brain injury; GCS, Glasgow Coma Scale; GCS-E, Extended Glasgow Coma Scale; IU, international unit; IV, intravenous; RCT, randomized controlled trial; SC, subcutaneously; DAI = diffuse axonal injury; DRS = Disability Rating Scale; FU = follow-up; ICP = intracranial pressure; ICU = intensive care unit; NR = not reported; NSE = neuron specific enolase; rhEPO = recombinant human EPO. |
*Age was reported as median (range) in the studies of Nichol et al., 2015, and Robertson et al., 2014; age was reported as mean ± SD in the studies of Bai et al 2018, Li et al 2016, Nirula et al., 2010, Aloizos et al., 2015, and Abrishamkar et al., 2012.. |
Of these seven included RCTs, six studies utilized a normal saline placebo control and one study did not mention the use of placebos18. The mean age of participants ranged from 26.3 to 43.8 years. The time from trauma to intervention was within 6 to 24 hours; different EPO regimens and dosages were used: five studies injected EPO, rhEPO, or epoetin alfa subcutaneously while two gave EPO intravenously with the total dosage ranging from 12,000 international unit(IU) to 120,000 IU, and the EPO administration mostly started within six hours except one started in 24 hours19. In our study, Nichol et al. set a 24-h time window for recruitment, which was at odds with other included RCTs. The major characteristics of the seven RCTs are shown in Table 1, with slightly different enrollment criteria of each. 20
Quality Assessment
Our assessments of each study’s risk of bias are summarized in Fig. 2. One trial was determined to be at high risk for bias18, two trials were at low risk for bias19 21, and the other four were at unclear risk of bias19 22−24. Four studies describe their methods for generating the randomization sequence and for allocation concealment. Five studies reported blinding of participants and investigators, which might induce performance bias.
Meta-analysis outcomes
Mortality
All studies reported on mortality (Fig. 3). Compared to the placebo arm, there was a no significant reduction in acute hospital mortality (EPO:7.1%;placebo:9.2%,RR = 0.71, 95% CI [0.32–1.58]; P = 0.41) and short-term mortality (RR = 0.64, 95% CI [0.26–1.60]; P = 0.34) in the patients with TBI (RR = 0.69, 95% CI [0.50–0.95]; p = 0.03). Our analysis of mid-term mortality(six months follow-up) included a total of 792 patients (74.5% of all included patients, drawn from the Robertson et al. 2014 and Nichol et al. 2015 studies). Here, the EPO-treated group showed a significantly lower mortality (EPO:11.2%,placebo:16.4%,RR = 0.69, 95% CI [ 0.48–0.98]; p = 0.04), the mid-term result of EPO on mortality was in accordance the overall results.
Figure 3 Effect of EPO intervention on mortality compared with control treatment at varying lengths of follow-up. Results are shown using a fixed-effect model with Risk ratio and 95% CIs. CI, confidence interval; M-H, Mantel-Haenszel.
Efficacy of EPO on Neurological Recovery
The neurological outcomes of TBI patients were evaluated by GOS or GOS-E varying from 10 weeks to 6 months (Fig. 4). only Li et al had showed EPO significant better control group. However, EPO was not associated with favorable neurological function improvement GOS-E 5–8 (RR = 1.22, 95%CI [0.82–1.81]; P = 0.33).
Safety outcomes
Thrombotic and cardiovascular events
Five studies reported the incidence of thromboembolic events15,17−19 at the end of follow-up (N = 1980). Compared to the placebo treatment, EPO therapy did not increase the incidence of total thromboembolic events (EPO:19.9%,placebo:22.2%,RR = 1.01, 95% CI [0.65–1.57]; P = 0.97) in TBI patients (Fig. 5). Our analysis of the pooled results demonstrate that there was no difference in rates of deep venous thrombosis (EPO:13.6,placebo:14.1%,RR = 0.98, 95% CI [0.45–2.14]; P = 0.96), pulmonary embolism (EPO:4.1%,placebo:3.3%,RR = 1.23, 95% CI [0.63–2.40]; P = 0.54), cardiac arrest (EPO:2.5%,placebo:1.8%,RR = 1.39, 95% CI [0.53–3.61]; P = 0.50), or myocardial infarction (EPO:0.9%,placebo:0.7%,RR = 1.22, 95% CI [0.29–5.07]; P = 0.79), (Fig. 5).
Other associated adverse events
Our pooled results showed no differences regarding other adverse events between the EPO and control groups (detail shown in Fig. 6). There were no differences in rates of pneumonia (EPO:10.8%,placebo:8.7%,RR = 1.25, 95%CI [0.73–2.15]; P = 0.42), sepsis (EPO:4.6%,placebo:3.5%,RR = 1.30, 95%CI [0.72–2.32]; P = 0.38), RBC transfusion (EPO:43.2%,placebo:45.7%,RR = 0.94, 95%CI [0.81–1.10]; P = 0.46), seizure (EPO:4.5%,placebo:3.7%,RR = 1.20, 95%CI [0.64–2.25]; P = 0.44), gastrointestinal complications (EPO:8.4%,placebo:7.5%,RR = 1.10, 95%CI [0.74–1.64]; P = 0.63). (Fig. 6).
Publication bias
A funnel plot did not reveal any obvious asymmetry (Fig. 7), and clear evidence of publication bias was not detected by Egger’s test.