Editorial 

October 9, 2024

Shifting Balance of the Risk-Benefit of Restrictive Transfusion Strategies in Neurocritically Ill Patients—Is Less Still More?

Alexis F. Turgeon, François Lauzier

JAMA. Published online October 9, 2024. doi:10.1001/jama.2024.20416

Acute neurological injuries, especially traumatic brain injuries and subarachnoid hemorrhages, often affect young and previously healthy patients with years of active life to come. Mortality remains very high following any severe acute brain injury, and importantly, a significant proportion of survivors have long-term sequelae that prevent them from independent living.¹ Despite proliferation of dedicated neurocritical care units and improvements in patient management over past decades, very few interventions clearly improve patient-oriented, clinically significant long-term functional outcomes.² Any interventions that could improve functional status in these patients may have a substantial impact.

The brain is particularly vulnerable to variations in oxygen delivery. While a reduction in oxygen delivery to other organs can be offset by increasing cardiac output, this compensation does not necessarily apply to the brain, where cerebrovascular autoregulation enacts changes in cerebral blood flow to maintain balance between oxygen delivery and consumption. However, these autoregulatory mechanisms are often impaired in acute brain injuries.³ As a result, oxygen content, determined by hemoglobin levels and oxygen saturation, plays a much greater role in these situations.^4,5 Given these deficient cerebral autoregulation mechanisms, the potential impact of anemia in neurocritically ill patients is a major concern.

Twenty-five years ago, the publication of the TRICC trial,⁶ which showed that a restrictive transfusion strategy was not inferior to a liberal strategy regarding 30-day mortality in critically ill adults, led to a dramatic change in transfusion practice, with restrictive thresholds being widely adopted to avoid inappropriate use of a scarce resource. However, very few neurocritically ill patients were enrolled in the TRICC trial, and long-term functional outcomes—more relevant to these patients⁷ than mortality and more impactful for clinical pratice⁸—were not assessed. Concerns were raised that even small decreases in hemoglobin could impair oxygen delivery and worsen patient outcomes. Nevertheless, strategies like blood transfusion, while theoretically increasing oxygen content, come with their own risks, and the actual oxygen delivery to tissues may not be as easily improved as one might assume. Due to these competing views, practice evolved to be highly variable, with some favoring restrictive and others liberal approaches.⁹

In this issue of JAMA, Taccone and colleagues¹⁰ present the results of the TRAIN trial, a large-scale, multicenter, international randomized clinical trial comparing the effects of 2 transfusion strategies, a liberal transfusion strategy and a restrictive one, in critically ill adult patients with acute brain injury. This trial, conducted across 72 sites in 22 countries, enrolled 850 neurocritically ill adults with anemia and acute traumatic brain injury, subarachnoid hemorrhage, or intracerebral hemorrhage. Patients underwent transfusion to maintain a hemoglobin level greater than 9 g/dL (liberal strategy group) or greater than 7 g/dL (restrictive strategy group). The primary outcome was the risk of an unfavorable neurological functional outcome at 6 months, defined as a score ranging from 1 (death) to 5 (lower moderate disability) on the Glasgow Outcome Scale Extended (GOS-E). The results showed a significant decrease in the incidence of an unfavorable outcome with the use of a liberal transfusion strategy (unfavorable outcome among 62.6% of patients in the liberal strategy group and 72.6% in the restrictive strategy group). These findings were consistent across all subgroups and post hoc analyses. There was no effect on intensive care unit and hospital lengths of stay, organ failure at 28 days, or mortality.

The TRAIN trial studied a heterogeneous population of neurocritically ill adults, a methodological decision that can be viewed both as a strength and a weakness. While studying heterogeneous populations may enhance generalizability, one could argue that the 3 neurocritical care conditions—traumatic brain injury, subarachnoid hemorrhage, and intracerebral hemorrhage—included in the trial have distinct pathophysiologies, making it unlikely that all would benefit equally from a liberal transfusion strategy. However, a subgroup analysis based on type of brain injury showed consistent findings in terms of both the direction of effect and the effect size.

Another aspect of the TRAIN trial is the use of an uncommon definition of an unfavorable outcome (death to independence in daily living activities without return to work/school) instead of the common definition (GOS-E from 1 to 4, which represents death to dependence in daily living activities). In addition to the use of an uncommon definition, others may argue that the simple dichotomization of the GOS-E score should no longer be used, as it does not account for the whole ordinal scale. Recent trials have used either a sliding dichotomy or a proportional odds approach to better leverage the ordinal 8-point scale of the GOS-E. Notwithstanding, the TRAIN trial showed a benefit of the intervention using a less powerful analytic strategy—simple dichotomization—with similar results observed in a post hoc analysis using the more common dichotomization cutoff of the GOS-E.

An inherent and unavoidable limitation of this study is the lack of blinding of the intervention. The authors mitigated this potential bias by blinding the outcome assessors to the treatment assignment. In critically ill patients, treatment intensity—or the perception that a study intervention is beneficial or harmful—can influence decisions to continue or withdraw life-sustaining therapies.^11-13 In the TRAIN trial, information on the incidence of withdrawal of life-sustaining therapies was not collected. However, a potential imbalance of withdrawal of life-sustaining therapies between groups,¹⁴ which may have occurred considering the 2.7% difference in mortality between the 2 groups, cannot entirely be excluded.

Another limitation of the TRAIN trial is the 2 sample size adjustments during the conduct of the trial. This was justified by a slow enrollment rate and the inability to complete the trial in a timely manner, and it highlights the challenges of conducting large-scale trials in critically ill patients with neurological injuries. Successfully conducting practice-changing trials in neurocritical care require more sites than trials studying more common critical conditions like sepsis or acute respiratory distress syndrome. The emergence of global collaborative research networks and the development of platform trials such as BRAINapt,¹⁵ an international adaptive platform trial for traumatic brain injury, will help increase research capacity and improve the conduct of practice-changing trials in this population.

The main findings of the TRAIN trial are consistent with those of the HEMOTION trial,¹⁶ a large-scale, multicenter international trial conducted in adults with moderate or severe traumatic brain injury and anemia. Although the observed benefit of a liberal transfusion strategy using a sliding dichotomy of the GOS-E in the HEMOTION trial was not statistically significant, both the primary analysis and all sensitivity analyses of the GOS-E pointed toward a benefit of the liberal strategy. This was also reflected in secondary outcome analyses, including functional independence and quality of life. The 2 trials differ in several aspects. TRAIN included a mixed population, studied a lower liberal transfusion threshold (9 g/dL), and used brain injury severity at randomization as an inclusion criterion. HEMOTION included solely patients with traumatic brain injury, used a higher liberal transfusion threshold (10 g/dL), and assessed brain injury severity in the emergency department following stabilization. Given the heterogeneous neurocritically ill population in TRAIN and the fact that most patients had traumatic brain injuries, some uncertainty may remain regarding the effectiveness of the liberal strategy in the less well-represented subpopulations with subarachnoid hemorrhage and intracerebral hemorrhage despite the apparent lack of differential effect in these subgroups. The upcoming results of the recently completed SAHaRA trial (NCT03309579), a multicenter transfusion trigger trial in adults with anemia and acute subarachnoid hemorrhage, will provide additional information on this specific patient population that was underrepresented in the TRAIN trial.

In the era of precision medicine, one might think that more specific and personalized triggers for red blood cell transfusions, such as invasive and noninvasive measures of cerebral oxygenation with parenchymal probes or near-infrared spectroscopy, could be more useful in determining the need for transfusion. However, with the current new evidence, the key question is no longer whether invasive or noninvasive measures of cerebral oxygenation should be used, but rather whether there is added value in using multimodal monitoring to inform transfusion decisions, as liberal transfusion thresholds appear widely beneficial. A liberal transfusion strategy is a simple intervention that can be easily and rapidly implemented worldwide, including in low- and middle-income countries. These countries have high incidences of traumatic brain injury and¹⁷ may have limited access to advanced specialized care but easier access to blood products, making this approach particularly feasible.

One barrier to the use of liberal transfusion thresholds in neurocritically ill patients is the concern about potential complications associated with transfusions. Blood products are now safer than ever. In TRAIN and HEMOTION, the adverse event rates, including thromboembolic events, were comparable between groups, with no clinically significant impact on patient-relevant outcomes. Although potential adverse effects of blood products should not be minimized, they must be weighed against significant improvement of clinically relevant and patient-relevant outcomes.

Another implementation barrier is the impact on blood bank systems and concern for the capacity to increase the availability of blood transfusions when blood is a scarce resource and shortages are already being faced.^18,19Since the TRICC trial, the rationale for advocating a restrictive strategy has been to save precious resources without harming patients. Fortunately, the weight of sparing precious resources has now been lifted from the shoulders of neurocritically ill patients, in whom serious concerns about the safety of a restrictive strategy have been raised by the TRAIN and HEMOTION trials. Based on the best available evidence, it is prudent to advocate a liberal transfusion strategy for these neurocritically ill patients.