Indication and survival bias: Threats to resuscitative endovascular balloon occlusion of the aorta outcomes research
We read with interest the study by Aso et al.1 Their objective was to compare early mortality between resuscitative endovascular balloon occlusion of the aorta (REBOA) and resuscitative thoracotomy (RT) in trauma patients with hemorrhagic shock. They found that although lower in REBOA-treated patients, mortality did not significantly differ between compared groups. However, this results could be flawed by the presence of indication and survival bias, a fact that is consistent in comparative effectiveness REBOA published literature.1–3
REBOA is thought to emulate resuscitative thoracotomy with the advantage of being a less invasive procedure, and thus it has been advocated to improve survival among noncompressible torso hemorrhage (NCTH) patients. To date, two studies2,3 have reported the effect of REBOA on mortality, compared it with RT. These studies showed that REBOA is a safe and effective option to be used in NCTH patients. However, similar to the study by Aso et al,1 results from these studies could be comprised by the presence of indication and survival bias.
Indication bias occurs in observational studies of interventions when patients are classified by the nonrandomized intervention they received during their medical treatment ignoring the factors that influenced the need for intervention.4 Survival bias arises when comparing patient groups in which patients may die before treatment is initiated.4 Both are selection biases that, if present, can comprise validity of results.
In the reports by Aso et al, DuBose et al., and Abe et al,1–3 patients in the RT group were more like to present with significantly lower values of blood pressure. Furthermore, a significantly higher proportion of RT patients arrived in cardiac arrests or required closed cardiopulmonary resuscitation. Moreover, in the study by Abe et al.,3 REBOA-treated patients presented to the emergency department with a significantly higher probability of survival, compared with those patients that underwent RT. The fact that across studies, RT patients were more likely to present with worse levels of hemodynamic instability and lower probability of survival indicates a consistent and systematic risk of indication and survival bias within individual studies. The problem with the presence of bias is that it jeopardizes both internal and external validity.5 Bias is a systematic error that produces a lack of internal validity or incorrect assessment of the association between exposure and an effect in a target population. The absence of internal validity comprises external validity that refers to the generalization of the results observed in one population to others.
The presence of this kind of selection biases in REBOA outcomes research suggests an absence of clear indications for REBOA use. Furthermore, it poses the question if REBOA is a comparable intervention to RT, which is a last rescue effort in the agonal patient, or if it is a procedure aimed to prevent rapid hemodynamic collapse while endovascular or open procedures achieve definitive hemorrhage control.
To develop a full picture of REBOA effect on mortality among NCTH patients, additional studies will be needed that use strategies to reduce the risk of biased results. The most powerful strategy to mitigate the risk of selection bias is to design and conduct a randomized controlled trial, which is challenging and hard to perform especially in trauma scenarios. However, well-designed and conducted trials are essential to advance in the understanding of trauma care and trauma outcomes research.
Finally, beyond bias and noise, REBOA researchers should always remember Frank Butler’s laws, especially the one that states: “Evidence does not drive advances in trauma care. People do that.