Transfusion Strategy Shows No Mortality Difference in Cancer Patients With Sepsis

When a cancer patient develops sepsis in the intensive care unit, doctors face a recurring question with no clear answer: how much blood should they transfuse? The question sounds technical, but it carries real weight. Transfuse too little and the patient's organs may fail from oxygen deprivation. Transfuse too much and you risk complications of your own. A new study of 658 cancer patients with sepsis offers some clarity, though not the kind that settles the debate.

Researchers looked back at a decade of ICU admissions—from 2014 through 2023—tracking how physicians made transfusion decisions for patients who were both critically ill with infection and fighting cancer. They sorted patients into two camps based on the hemoglobin thresholds that triggered transfusions. The restrictive approach held back blood until hemoglobin dropped below 7 to 8 grams per deciliter. The liberal approach started transfusing at 9 to 10 grams per deciliter. The distinction mattered because it reflected fundamentally different philosophies about how much blood a failing body actually needs.

What the researchers found was striking in its evenness. After adjusting for the differences between the two groups—a statistical technique called inverse probability weighting designed to account for the fact that sicker patients tended to fall into the restrictive camp—there was no meaningful difference in mortality. At 28 days, both groups fared similarly. At 90 days, the raw numbers suggested the restrictive group did worse, but once the researchers accounted for how much sicker those patients were at baseline, the gap vanished. The study included 658 patients total, and neither transfusion strategy proved superior at keeping people alive.

But mortality was not the whole story. The restrictive approach came with a cost that persisted even after statistical adjustment: patients managed with lower transfusion thresholds developed acute kidney injury more often than those transfused more liberally. This association held up under scrutiny, suggesting it was not merely a statistical artifact. Yet the researchers were careful to note that association is not causation. They could not prove that withholding blood damaged kidneys; they could only show that the two things occurred together more often in the restrictive group.

The baseline characteristics revealed why caution was warranted. Patients in the restrictive group were sicker to begin with, as measured by standard severity scores. They had higher APACHE II and SOFA scores, the numerical tools intensivists use to gauge how close a patient is to the edge. The two groups also differed in their cancer types: lung cancer patients clustered in the liberal transfusion group, while gastrointestinal malignancies predominated in the restrictive group. These imbalances meant that even with statistical adjustment, the groups were not truly comparable. A patient with advanced lung cancer facing sepsis is not the same as a patient with gastric cancer in the same crisis.

The study's design—retrospective and observational rather than a controlled trial—meant that physicians had chosen the transfusion strategy based on their own judgment and experience. This introduced what researchers call selection bias: the sickest patients may have been managed restrictively precisely because they were sicker, or because their particular cancers or organ function suggested a conservative approach. The fact that mechanical ventilation duration, vasopressor use, and dialysis requirements did not differ between groups suggested the overall intensity of care was comparable, but it did not erase the fundamental uncertainty about whether the transfusion strategy itself caused the kidney injury or merely accompanied it.

What emerges is a study that answers one question while raising others. In cancer patients with sepsis, the choice between restrictive and liberal transfusion does not appear to determine whether patients live or die. But it may shape how their organs behave in the interim. The researchers called for prospective randomized trials—the gold standard of medical evidence—to test whether individualized transfusion strategies might work better than either blanket approach. Until then, the question remains open, and physicians will continue to make these decisions in the fog of incomplete knowledge, guided by severity scores and instinct.