Cause and Timing of Death and Subgroup Differential Effects of Erythropoietin in the EPO-TBI Study.
Skrifvars MB., French C., Bailey M., Presneill J., Nichol A., Little L., Durantea J., Huet O., Haddad S., Arabi Y., McArthur C., Cooper DJ., Bellomo R.
The EPO-TBI study randomized 606 patients with moderate or severe traumatic brain injury (TBI) to be treated with weekly epoetin alfa (EPO) or placebo. Six month mortality was lower in EPO treated patients in an analysis adjusting for TBI severity. Knowledge of possible differential effects by TBI injury subtype and acute neurosurgical treatment as well as timing and cause of death (COD) will facilitate the design of future interventional TBI trials. We defined COD as cerebral (brain death, cerebral death with withdrawal, or death during maximal care) and non-cerebral (death following withdrawal or during maximal care, which had a non-cerebral cause). The study included 305 patients treated with EPO and 297 treated with placebo, with COD recorded in 77 (99%) out of 78 non-survivors. Median time to death in patients dying of cerebral COD was 8 days (interquartile range [IQR] 5-16) compared with 29 days (IQR 7-56) (p = 0.01) for non-cerebral COD. When assessing subgroups by admission CT scan injury findings, we found no significant differential effects of EPO compared with placebo. However, EPO appeared more effective in patients with an injury type not requiring a neurosurgical operation prior to intensive care unit (ICU) admission (odds ratio [OR] 0.29, 95% confidence interval [CI] 0.14-0.61, p = 0.001, p for interaction = 0.003) and in this subgroup, fewer patients died of cerebral causes in the EPO than in the placebo group (5% compared with 14%, p = 0.03). In conclusion, most TBI deaths were from cerebral causes that occurred during the first 2 weeks, and were related to withdrawal of care. EPO appeared to specifically reduce cerebral deaths in the important subgroup of patients with a diffuse type of injury not requiring a neurosurgical intervention prior to randomization.