Each publication from Maitland, et al., gives us a glimpse of a greater medical truth, but our view is obstructed by questions about external validity.
This group published the landmark FEAST trial that questioned the safety and utility of one of the most ubiquitous interventions in medicine today, the fluid bolus. (N Engl J Med. 2011;364:2483; http://bit.ly/2L0ozdO.) The authors enrolled 3141 pediatric sepsis patients at six clinical centers in Africa, and patients were randomized to receive an IV fluid bolus (saline or albumin) plus a continuous infusion or infusion alone.
The results were surprising, to say the least. The trial was stopped premature of its 3600 predetermined sample size because of an increased rate of 48-hour mortality, 10.6 percent, 10.5 percent, and 7.3 percent in the albumin-bolus, saline-bolus, and no bolus control groups, respectively. This was the first trial empirically examining the efficacy of a fluid bolus for septic shock, and the authors found that IV fluid boluses were associated with increased 48-hour and 28-day mortality.
Yet, eight years later, rapid fluid administration is still paramount in managing septic shock. Many argue that these results may not be externally valid outside the context in which they were tested. Are fluid boluses detrimental when used in critically ill adults with access to ICU care? I previously wrote about the growing body of evidence suggesting Maitland's results are applicable more broadly than originally considered (EMN. 2019;41:1; http://bit.ly/2J46WKK), but concerns about the FEAST trial's external validity have limited its influence on managing sepsis and septic shock.
Now, the TRACT trial by Maitland, et al., asks us to consider the validity of another long-held medical practice. (N Engl J Med. 2019;381:407; http://bit.ly/2KT2aPi.) Much like administering IV crystalloid, the current management of severe anemia is, in theory, fairly straightforward. Maitland was guided by the TRICC trial, in which critically ill ICU patients with no signs of active hemorrhage were randomized to different transfusion thresholds, 7 g/dL or 10 g/dL. (N Engl J Med. 1999;34:409; http://bit.ly/2MvbbS9.) The authors enrolled 838 patients, 413 in the restrictive group and 420 in the liberal group. Patients in the liberal group successfully had their hemoglobin maintained at a higher level than the restrictive group. The average daily hemoglobin concentrations were 8.5±0.7 g/dL in the restrictive group and 10.7±0.7 g/dL in the liberal one. Clinicians were required to transfuse significantly larger quantities of PRBCs: 2.6±4.1 PRBCs per patient in the restrictive group v. 5.6±5.3 in the liberal group to achieve this higher hemoglobin value.
Thirty-day mortality was 18.7 percent in the restrictive group v. 23.3 percent in the liberal group (p=0.11). In-hospital mortality was 22.2 percent v. 28.1 percent in the restrictive and liberal groups, respectively (p=0.05). ICU mortality and 60-day mortality trended toward the restrictive group. Cardiac events (MI, pulmonary edema) were also more frequent in the liberal group (13.2% v. 21.1%).
The results of the TRICC trial have been validated in many pediatric and adult populations. (N Engl J Med. 2007;356:1609, http://bit.ly/31Tk7UN; N Engl J Med. 2013;368:11, http://bit.ly/2Npa2ek; N Engl J Med. 2017;377:2133, http://bit.ly/2ZkTzKu.) It is clear an empirically restrictive strategy is superior to an empirically liberal one, but this trial demonstrated 7 g/dL was preferable to 10 g/dL, not that 7 g/dL was better than 6 g/dL and 5 g/dL and not whether empiric transfusion strategies based on hemoglobin levels were superior to transfusion triggers based on signs indicating the anemia was leading to physiological distress. These questions have been relegated to the musings of the rebellious mind, but the TRACT trial provides clinical data suggesting a single threshold may not be ideal. (N Engl J Med. 2019;381:407; http://bit.ly/2KT2aPi.)
Maitland, et al., enrolled children 2 months to 12 years old with uncomplicated severe anemia in a factorial, open-label, randomized, controlled trial. The patients were randomized to immediate transfusion or a control group in which a transfusion was deferred unless the hemoglobin dropped below 4 g/dL or clinical signs of severe anemia developed (prostration or respiratory distress). Patients randomized to immediate transfusion received 20 mL/kg or 30 mL/kg of whole-blood equivalent determined by a second randomization. (Those results are in N Engl J Med. 2019;381:420; http://bit.ly/2Zj26NW.) If a second transfusion was indicated, it was done at the same volume as the first. Patients randomized to the control group who met the threshold for transfusion were transfused 20 mL/kg of whole-blood equivalents.
The Maitland group enrolled 1565 children, 778 to immediate transfusion (390 in the 20 mL group, 388 in the 30 mL) and 787 as controls. All patients in the immediate transfusion group received at least one transfusion, but only 386 (49.0%) in the control group did. The mean total volume of whole-blood equivalent transfused per child during initial hospitalization was 314±228 mL v. 142±224 mL in the immediate-transfusion and control groups, respectively.
The more aggressive transfusion strategy resulted in fewer patients experiencing a severe anemia (hemoglobin<4g/dL; 1.4% v. 39.3%) and more children experiencing an early hemoglobin recovery (>9 g/dL; 51.3% v. 5.5%). But no difference in primary outcome, 28-day mortality, was seen: 0.9 percent in the immediate group v. 1.7 percent in the control group. Nor did they identify differences in mortality at three or six months or in complications from severe anemia or readmission to the hospital during the subsequent six months. In fact, the difference in hemoglobin levels between the two groups at 28 days was minimal, only 0.6 g/dL higher in the immediate-transfusion group.
Once again, Maitland, et al., have challenged what we thought we knew about medicine. The authors generated data that can be applied instantly by many practicing around the world. We are uncertain how these results apply to patients outside the population studied, which brings us back to external validity: Can a highly restrictive transfusion strategy be employed in patients outside the trial hospitals in Uganda and Malawi? Is Maitland's population unique, preventing these results from being applied to a broader population of anemic children or even adults? Like the FEAST trial, 62.9 percent had malaria and 21.7 percent had sickle cell disease. Should this limit the trial's generalizability? These concerns remain unanswered without further study, but almost every study examining transfusion strategies in various patients have found restrictive strategies preferable to liberal ones.
Can we apply the results to a population with access to ICU care? The authors' major concern was whether allowing hemoglobin levels to fall so low without intensive monitoring was safe, which was why a strict hemoglobin monitoring strategy was incorporated into the methodology. Frequent blood draws are well within the capabilities of any modern ICU, so can we apply the results of this trial to those without the capability to monitor hemoglobin levels regularly?
Finally, can these results be applied to adults with critical anemia? Current pediatric transfusion guidelines appear fairly similar to adult ones (Pediatr Crit Care Med. 2018;19[Suppl 1]:S98), recommending a transfusion trigger of 7 g/dL. The TRIPICU trial in a sense validated the TRICC trial findings in children, demonstrating that a transfusion trigger of 7 g/dL was superior to one of 9.5 g/dL. (N Engl J Med. 2007;356:1609; http://bit.ly/31Tk7UN.) But like the TRICC trial, the authors only asked a dichotomous question and did not examine whether a lower transfusion threshold was preferred to either level.
A greater medical truth hides in the pages of this document. Transfusion of red blood cells can likely occur at a much lower threshold than what we generally employ, and the level at which anemia becomes detrimental probably varies from patient to patient. It is doubtful these results, like FEAST, will be broadly applied. Instead, we are stuck in an endless cycle of dismissal and denial. A trial like this could not have been done outside of this population, but we are unable to apply the results generally because of that population. It is unclear whether the TRACT trial results are generalizable to blood transfusion strategies outside this cohort.
The results should make us question the absurd dichotomy in which we currently exist, where a patient with a 7.1 g/dL hemoglobin is fine and a patient at 6.9 g/dL is in urgent need of red blood cells, with no consideration of the clinical context surrounding these abstract data points.
Dr. Spiegelis an assistant professor of emergency medicine and critical care at Washington Hospital Center in Washington, D.C. Visit his blog athttp://emnerd.com, follow him on Twitter@emnerd_, and read his past articles athttp://bit.ly/EMN-MythsinEM.