Invited Commentary
Critical Care Medicine
March 28, 2025
Active Prevention and Treatment of Post–Critical Illness Anemia—Is It Time for Large Effectiveness Trials?
Timothy S. Walsh
JAMA Netw Open. 2025;8(3):e252368. doi:10.1001/jamanetworkopen.2025.2368
Anemia is prevalent following critical illness and is associated with delayed recovery, functional impairments, and reduced quality of life.1 Evidence tells us that restrictive transfusion practice in the intensive care unit (ICU) is at least as effective as liberal transfusions for most patients. One consequence of restrictive transfusion is a high prevalence of post–critical illness anemia, which can persist for months.1,2 There is remarkably little high-quality research exploring whether interventions to prevent anemia and/or promote its recovery are effective for increasing hemoglobin concentration, and whether these interventions benefit patients’ health. Warner and colleagues3 report on a randomized clinical trial that advances our knowledge of this issue and supports the need for further research.
In their trial, Warner and colleagues3 randomized 100 critically ill adult patients with moderate to severe anemia (defined as hemoglobin level <10g/dL) to receive an intervention bundle comprising optimized phlebotomy practices, clinical decision support, and pharmacological treatment with intravenous (IV) iron. The comparator group received standard care. The population mainly consisted of surgical patients (65.0% admitted after surgery, of whom 71.0% underwent cardiac surgery), with 76.0% managed in a surgical ICU. The authors found a significant improvement in the primary outcome: hemoglobin concentration 1 month after hospital discharge. Median (IQR) hemoglobin concentrations were 12.2 (11.8-13.0) g/dL (to convert to g/L, multiply by 10.0) in the intervention group and 11.5 (10.2-12.6) g/dL in the standard care group, with an adjusted mean difference of 0.69 (95% CI, 0.13-1.20) g/dL. This difference was not observed at earlier time points—namely, at ICU and hospital discharge—but was maintained at 3 months, when the mean difference was 0.58 (0.02-1.10) g/dL.
This trial was well conducted, powered at 80% to detect a clinically relevant difference in hemoglobin concentration of 1 g/dL at 1-month posthospital discharge, with conservative estimates for dropout rates due to death or loss to follow-up (26 of 100 patients). As only 3 patients (2 [4.1%] in the intervention group; 1 [2.0%] in the standard care group) died during the study, and only 12 were lost to follow-up at 1 month, the power of the trial was effectively over 80%. The trial minimized bias through robust randomization concealment and blinding of outcome assessors, although none of the intervention components were blinded from clinicians and participants, introducing potential performance bias. Major risks from lack of blinding likely included differences in red blood cell (RBC) transfusion practices or crossover in IV iron use in the standard care group. However, the authors followed their local evidence-based RBC transfusion guidelines during hospitalization, and only 7 patients (1 [2.0%] in the intervention group; 6 [11.7%] in the standard care group) received postdischarge RBC transfusions. Only 6 patients (11.7% in the standard care group) received IV iron later during hospitalization as part of standard care. It seems unlikely that these factors were major confounders to the observed effect, and any bias would most likely have deflated rather than inflated the observed difference in the primary outcome.
Another strength was the preplanned randomization stratification by relevant baseline anemia cause (iron responsiveness vs nonresponsiveness) and by ICU admission category (nonsurgical vs surgical), along with adjustment of the primary outcome analysis for baseline factors that might influence the outcome, including age, sex, baseline hemoglobin level, and admission type. This adjustment is potentially important in relatively small unblinded trials, where random baseline imbalances in potential effect moderators can affect the observed effect size.
As this trial tested the effectiveness of a bundle of care, it is relevant to try to understand which of the 3 components of this bundle was most important. To do so, we need to unpack how the care bundle actually changed the treatment given. This is often called a process evaluation and is especially important in understanding the active ingredients in complex health care interventions.4 The authors provide some useful information about the delivery of the 3 care bundle components. First, the phlebotomy optimization team decreased the median (IQR) number of blood draws from 46 (20-74) to 32 (24-55), and the median (IQR) total volume drawn from 142 (68-195) mL to 32 (22-54) mL. Second, virtually all patients were classed as iron responsive according to the prespecified definition, and 98.0% received the full predefined IV iron dose. No patients received erythropoietin, but this was only planned for iron-nonresponsive anemia. Treatment changes related to the third component (namely, clinician decision support with alerts in the electronic health record and daily rounds to optimize other relevant care) were less clear as these were not systematically recorded. Given that the effect on hemoglobin concentration was modest and not statistically significant at hospital discharge, we can speculate that differences in phlebotomy volumes (which were modest) and clinical decision-making were the less-active ingredients in the bundle. The separation in hemoglobin concentration increased after hospital discharge and was maintained over 3 months, supporting the conjecture that IV iron may have contributed most to the effect on the primary outcome, with ongoing effects on erythropoiesis after discharge. This conjecture is supported by all patients fulfilling the iron-responsive definition at baseline and the high number of postoperative patients in the trial population whose anemia resulted in part from acute blood loss.
This observation raises potential limitations of the generalizability of the findings to patients in medical ICU or patients with more prolonged critical illness. In these cases, more severe and persisting inflammation may limit iron responsiveness even when delivered by IV to overcome the block to intestinal iron absorption that results from interleukin-6 mediated upregulation of hepcidin production, which is the key regulator of iron absorption and availability.5 A relationship between persisting systemic inflammation and anemia recovery after critical illness has been demonstrated previously.2
Hemoglobin concentration is a measure of oxygen-carrying capacity, but it is best considered as an intermediate outcome of a presumed causative pathway that links anemia to patient-centered outcomes. These outcomes include fatigue, exercise capacity, and quality of life. The prespecified secondary outcomes in the Warner and colleagues’3trial included validated measures of fatigue, exercise capacity, activities of daily living, cognitive function, anxiety and depression, and overall quality of life. It is reasonable to hypothesize that anemia correction may improve all of these measures. The authors found no statistically significant differences in secondary outcomes, but this is not surprising given the sample size. They appropriately reported effect sizes with 95% CIs, which allows an estimation of the range of effects that might result from the trial intervention if the trial were repeated in a larger, adequately powered study. The authors found consistent direction of effects favoring the intervention for fatigue scores, cognitive function, psychological well-being, and overall quality of life. These findings support the rationale for a future trial that is designed to detect improvements in patient-centered outcomes.
Other trials of iron therapy with or without erythropoietin support an effect on hemoglobin recovery following critical illness.6,7 The trial by Warner and colleagues3 provides further evidence that care bundles that minimize iatrogenic blood loss and promote anemia recovery are effective for increasing hemoglobin concentration compared with existing or laissez-faire approaches. We now need adequately powered trials to determine the clinical effectiveness of these interventions for patient-centered outcomes and evaluation of the economic benefits of implementing them at scale as part of best practice.
